MI VDF API

---

REVISION HISTORY

Revision No.	Description	Date
2.03	Initial release	11/12/2018
2.04	Document refinement	04/12//2019
2.05	Add Work Flow Fix MD param info	10/12/2020
2.06	Fix VG param info. Add description of new structure.	11/02/2020
2.07	Add Flowchart Title Add MI_MD_ResultSize_t struct Fix variables and funcs hyperlinks Add MI_OD_Result_t struct member Add proc debug info	11/10/2021
2.08	Change 16RstBufSize to u32RstBufSize	03/28/2022
2.09	Add 64bit physical address of MD/OD	04/12/2022
2.10	Add polygon region detection for VG	12/06/2023
2.11	Modify vdf debug information	03/10/2025
2.12	Modify document layout Add the introduction of Function, Application Scenario, Chip Difference and Development	11/21/2025

---

1. OVERVIEW

1.1. Module Description

MI_VDF is a device driver based on an algorithm library that implements motion detection (MD), occlusion detection (OD), and virtual fence/line detection (Virtual Gate, VG).

1.2. Basic structure

As a security detection functional module, VDF usually connects to the SCL at the front end. The monitored images are scaled and then input into the VDF for algorithm processing. The results are then output directly for user use, so the VDF does not need to connect to the back end.

A typical application scenario for VDF:

Figure 1-1 Typical Application Scenarios of VDF

As shown in Figure 1-1:

• The three input sources are processed through SCL scaling and other treatments to output three sets of data to the VDF for MD, OD, or VG detection. The detection results are directly fed back to the user.

1.3. Function Introduction

VDF supports the following functions:

1. Supports motion detection. When an object is detected and the alarm conditions are met, the alarm will be triggered.

2. Supports obstruction detection. When an object blocking the camera is detected and the alarm conditions are met, it triggers an alarm.

3. Supports virtual fences and line detection. When an object crosses a virtual fence or line and meets the alarm conditions, an alarm is triggered.

4. Set and get channel parameters (algorithm static and dynamic parameters).

5. Supports enabling single/multiple channels. The operating mode can be a single mode or a combination of multiple modes (MD, OD, and VG).

6. Supports dynamically disabling or enabling a specific mode.

7. Supports dynamically disabling or enabling a specific channel.

1.4 Application scenarios

The VDF module can be applied to both Pure Linux systems and Pure RTOS or Dual OS systems.

1.5 Chip differences

The VDF module is a purely software module, with no differences across various chips.

1.6 Working principle

1.6.1 MD Principle Description

The MD algorithm is divided into three modes (foreground mode, SAD mode, and difference mode)

Foreground Mode (MDALG_MODE_FG): For each pixel modeling, the current Y value is estimated to fall on the foreground or background zone of the model, so as to determine whether the pixel is a foreground spot or a background point.

SAD mode (MDALG_MODE_SAD): calculates the difference between the current frame and the background frame based on the block size to determine whether the block is a dynamic or static block. The background frame is initialized as the first frame, and then the pixel is used as a unit to determine whether this pixel is a foreground point or a background point, and different weights are given to update the background frame.

Frame Difference Mode (MDALG_MODE_FRAMEDIFF): Calculates the difference between the current frame and the background frame based on the block size to determine whether the block is a dynamic or static block.

Note: Processing speed: SAD mode > Frame Difference mode > Foreground mode. Background blending: Frame Difference mode has no background blending, SAD and FG have background blending. SAD mode is recommended.

1.6.2 OD Principle Description

First, the background feature frame is created with M frame (120 frames recommended), and then converted to a binary image (binary image) as the feature frame. The feature binary image of the current frame is compared with the feature binary image of the current background, and the multi-layer threshold is determined to determine whether to trigger the alarm.

1.6.3 VG Principle Description

After VG is activated, you need to let the screen stand still for 3 seconds to create a background. Then, the difference between the features of the current frame and the background frame is judged, and if the difference is greater than the threshold, it is judged to be the foreground. Finally, it is judged whether the foreground object crosses the set alarm line segment and area, and whether the directionality is satisfied, and then decides whether to make an alarm.

1.7 Development process

1.7.1 Compilation Configuration

1. Enter the alkaid project root directory, make menuconfig.

2. Press the Enter key to enter the Generic Options sub-option.

   

3. Press the Enter key to enter the Interface Compile Config sub-option.

   

4. Select the VDF module with the space key and recompile the entire image.

   

1.7.2 API Call

The interface call sequence of the VDF module is as follows:

Figure 1-2: Function Call Process of the VDF Channel

The above figure is the call process of opening a VDF, but VDF can support multiple channels, and each channel can be set to one of MD/OD/VG working modes.

Each VDF channel must be controlled by itself to call MI_VDF_CreateChn, MI_VDF_SetChnAttr, MI_VDF_EnableSubWindow, MI_VDF_GetResult, MI_VDF_PutResult, MI_VDF_DestroyChn independently.

1.8. Example Introduction

1.8.1 vdf-related code

Here, MD is used as an example for illustration. For VG, the interface call sequence can refer to Figure 1-1 in section 1.7.2 for setting up VDF. The following provides an example process for multiple channels. Two VDF channels are enabled, with channel 0 set to MD mode and channel 1 set to OD mode. This example uses motion detection for demonstration and can be followed according to the steps below.

1. Set MD's static/dynamic parameters ---> Create channel ---> Enable channel

2. User layer obtains computation results ---> Return result buffer

3. Unregister channel

1.8.1.1 Build vdf module

#include <threads.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <assert.h>
#include <signal.h>

#include <mi_sys_datatype.h>
#include <mi_sys.h>
#include <mi_vdf_datatype.h>
#include <mi_vdf.h>

MI_VDF_Init();

/* Turn on the MD mode master switch */
MI_VDF_Run(E_MI_VDF_WORK_MODE_MD);

MI_VDF_CHANNEL vdfChn = 0;
MI_VDF_ChnAttr_t stAttr;
MI_VDF_MdAttr_t stMdAttr;
int width = 640, height = 360;

/* Set static parameters */
stMdAttr.stMdStaticParamsIn.width = width;
stMdAttr.stMdStaticParamsIn.height = height;
stMdAttr.stMdStaticParamsIn.roi_md.num = 4;
stMdAttr.stMdStaticParamsIn.md_alg_mode = MDALG_MODE_SAD;
stMdAttr.stMdStaticParamsIn.mb_size = MDMB_MODE_MB_8x8;
stMdAttr.stMdStaticParamsIn.sad_out_ctrl = MDSAD_OUT_CTRL_8BIT_SAD;
stMdAttr.stMdStaticParamsIn.color = 1;

/* Effective Monitoring Range Coordinates */
stMdAttr.stMdStaticParamsIn.roi_md.pnt[0].x = 0;
stMdAttr.stMdStaticParamsIn.roi_md.pnt[0].y = 0;
stMdAttr.stMdStaticParamsIn.roi_md.pnt[1].x = width - 1;
stMdAttr.stMdStaticParamsIn.roi_md.pnt[1].y = 0;
stMdAttr.stMdStaticParamsIn.roi_md.pnt[2].x = width - 1;
stMdAttr.stMdStaticParamsIn.roi_md.pnt[2].y = height - 1;
stMdAttr.stMdStaticParamsIn.roi_md.pnt[3].x = 0;
stMdAttr.stMdStaticParamsIn.roi_md.pnt[3].y = height - 1;

/* Set dynamic parameters */
stMdAttr.stMdDynamicParamsIn.sensitivity = 80;
stMdAttr.stMdDynamicParamsIn.learn_rate = 2000;
stMdAttr.stMdDynamicParamsIn.md_thr = 16;
stMdAttr.stMdDynamicParamsIn.obj_num_max = 0;

stMdAttr.u8MdBufCnt = 4;
stMdAttr.u8VDFIntvl = 1;
stMdAttr.ccl_ctrl.u16InitAreaThr = 8;
stMdAttr.ccl_ctrl.u16Step = 2;
stMdAttr.u8Enable      = 1

stAttr.stMdAttr = stMdAttr;
stAttr.enWorkMode = E_MI_VDF_WORK_MODE_MD;

/* Create a channel, this interface will set the relevant static and dynamic parameters */
ret = MI_VDF_CreateChn(vdfChn, &stAttr);
if (ret != MI_SUCCESS)
{
    ST_DBG("MI_VDF_CreateChn ret=0x%x\n", ret);
}

/* enable this channel, start working normally*/
MI_VDF_EnableSubWindow(vdfChn, 0, 0, 1);

1.8.1.2 Get the calculation result

/* Obtain monitoring results */
MI_MD_static_param_t *pstMdStaticParamsIn = NULL;
MI_U8 *stSadDataArry = NULL;
MI_U8 *fgDataArry = NULL;
int width,height,col,row;
int buffer_size;

pstMdStaticParamsIn = (MI_MD_static_param_t *)&stAttr.stMdAttr.stMdStaticParamsIn;
width = pstMdStaticParamsIn->width;
height = pstMdStaticParamsIn->height;

if (pstMdStaticParamsIn->mb_size == MDMB_MODE_MB_4x4)
{
    col = width >> 2;
    row = height >> 2;
}
else if (pstMdStaticParamsIn->mb_size == MDMB_MODE_MB_8x8)
{
    col = width >> 3;    // 40
    row = height >> 3;    // 22
}
else    // MDMB_MODE_MB_16x16
{
    col = width >> 4;
    row = height >> 4;
}

buffer_size = col * row;    // MDSAD_OUT_CTRL_8BIT_SAD
if (pstMdStaticParamsIn->sad_out_ctrl == MDSAD_OUT_CTRL_16BIT_SAD)
{
    buffer_size *= 2;
}
if (stAttr.stMdAttr.stMdStaticParamsIn.md_alg_mode == MDALG_MODE_SAD)
{
    stSadDataArry = (MI_U8 *)malloc(buffer_size + 1);
    if (NULL == stSadDataArry)
    {
        ST_ERR("can not malloc stSadDataArry buf.\n");
        goto exit;
    }
    memset(stSadDataArry, 0x0, buffer_size + 1);
    ST_DBG("enter md task loop, md model is %d\n", MDALG_MODE_SAD);
}
if (stAttr.stMdAttr.stMdStaticParamsIn.md_alg_mode == MDALG_MODE_FG)
{
    fgDataArry = (MI_U8 *)malloc(col * row + 1);
    if (NULL == fgDataArry)
    {
        ST_ERR("can not malloc fgDataArry buf.\n");
        goto exit;
    }
    memset(fgDataArry, 0x0, col * row + 1);
    ST_DBG("enter md task loop, md model is %d\n", MDALG_MODE_FG);
}

MI_VDF_Result_t stVdfResult = { (MI_VDF_WorkMode_e)0 };
MI_U8 *pu8MdRstData = NULL;

/* Get the output buffer, which stores the calculation results */
ret = MI_VDF_GetResult(vdfChn, &stVdfResult, 0);
if (MI_SUCCESS == ret)
{
  if (1 == stVdfResult.stMdResult.u8Enable)
  {
      if (stVdfResult.stMdResult.pstMdResultSad != NULL &&
              stVdfResult.stMdResult.pstMdResultSad->enOutCtrl == MDSAD_OUT_CTRL_8BIT_SAD &&
              stAttr.stMdAttr.stMdStaticParamsIn.md_alg_mode == MDALG_MODE_SAD)
      {
        pu8MdRstData = (MI_U8 *)stVdfResult.stMdResult.pstMdResultSad->paddr;
        memcpy(stSadDataArry, pu8MdRstData, buffer_size);
        sMdSADFlag = 1;
      }
      else if (stVdfResult.stMdResult.pstMdResultObj != NULL &&
                stAttr.stMdAttr.stMdStaticParamsIn.md_alg_mode == MDALG_MODE_FG)
      {
        // value = 0 or 255
        memcpy(fgDataArry, stVdfResult.stMdResult.pstMdResultStatus->paddr, col * row);
        sMdFGFlag = 1;
      }

      if (sMdSADFlag && stAttr.stMdAttr.stMdStaticParamsIn.md_alg_mode == MDALG_MODE_SAD)
      {
        /* Print the calculation results */
        printf("pu8MdRstData:===>\n");
        for (int i = 0; i < row; i++)
        {
            printf("===>");
            for (int j = 0; j < col; j++)
            {
                printf("%02d, ", stSadDataArry[i * col + j]);
            }
            printf("<===\n");
        }
        printf("<===\n");
      }

      if (sMdFGFlag && stAttr.stMdAttr.stMdStaticParamsIn.md_alg_mode == MDALG_MODE_FG)
      {
        /* Print the calculation results */
        printf("pu8MdRstData:===>\n");
        for (int i = 0; i < row; i++)
        {
            printf("===>");
            for (int j = 0; j < col; j++)
            {
                printf("%02d, ", fgDataArry[i * col + j]);
            }
            printf("<===\n");
        }
        printf("<===\n");
      }
  }

  /* Return output buffer */
  MI_VDF_PutResult(vdfChn, &stVdfResult);
}

1.8.1.3 Destruct vdf module

/* diable this channel, start working normally*/
MI_VDF_EnableSubWindow(vdfChn, 0, 0, 0);

/* Turn off the MD mode master switch */
MI_VDF_Stop(E_MI_VDF_WORK_MODE_MD)；

/* Destroy this channel */
MI_VDF_DestroyChn(vdfChn)；
MI_VDF_Uninit()；

---

2. API REFERENCE

2.1. VDF API List

API Name	Function
MI_VDF_Init	Initialize MI_VDF module
MI_VDF_Uninit	Un-initialize MI_VDF Module
MI_VDF_CreateChn	Create channel for MD/OD/VG
MI_VDF_DestroyChn	Destroy created channel MD/OD/VG
MI_VDF_SetChnAttr	Set channel (MD/OD/VG) attribute
MI_VDF_GetChnAttr	Get attribute from channel (MD/OD/VG)
MI_VDF_EnableSubWindow	Enable sub windows for (MD/OD/VG) channel
MI_VDF_Run	Start detection for channel (MD/OD/VG)
MI_VDF_Stop	Stop detection for channel (MD/OD/VG)
MI_VDF_GetResult	Get detection result from channel (MD/OD/VG)
MI_VDF_PutResult	Release detection result from channel (MD/OD/VG)
MI_VDF_GetLibVersion	Get version number of the specified channel (MD/OD/VG)
MI_VDF_GetDebugInfo	Get debug info of the specified channel. (Only VG support this feature)

2.2. VDF API Description

2.2.1. MI_VDF_Init

• Functions

  Initialize the VDF module.

• Syntax

  MI_S32 MI_VDF_Init(void);
  

• Return Value

  • Zero: Successful

  • Non-zero: Failed, see error code for details

• Dependency

  • Header File: mi_sys.h, mi_md.h, mi_od.h, mi_vg.h, mi_vdf.h

  • Library File: libOD_LINUX.a, libMD_LINUX.a, libVG_LINUX.a, libmi_vdf.a

• Note

  • MI_VDF_Init needs to be called after calling MI_SYS_Init.

  • It is not possible to call MI_VDF_Init repeatedly.

• Related APIs

  MI_VDF_Uninit

2.2.2. MI_VDF_Uninit

• Functions

  Destructing the MI_VDF system, before calling MI_VDF_Uninit , you need to ensure that the created VDF channels have been disabled and the video detection results of all channels are released

• Syntax

  MI_S32 MI_VDF_Uninit (void);
  

• Return Value

  • Zero: Successful

  • Non-zero: Failed, see error code for details

• Dependency

  • Header File: mi_sys.h, mi_md.h, mi_od.h, mi_vg.h, mi_vdf.h

  • Library File: libOD_LINUX.a, libMD_LINUX.a, libVG_LINUX.a, libmi_vdf.a

• Note

  • Before the MI_VDF_Uninit call, you need to ensure that all created VDF channels are in the disable state.

  • Before the MI_VDF_Uninit call, you need to ensure that the results of all created VDF channels are released.

• Example

  See MI_VDF_Init For example.

2.2.3. MI_VDF_CreateChn

• Functions

  Create a video detection (MD/OD/VG) channel.

• Syntax

  MI_S32 MI_VDF_CreateChn(MI_VDF_CHANNEL VdfChn, const MI_VDF_ChnAttr_t* pstAttr);
  

• Parameters

  Parameter Name	Description	Input/Output
  VdfChn	Specify the video detection channel number.	Input
  pstAttr	Set the video channel property value.	Input

• Return Value

  • Zero: Successful

  • Non-zero: Failed, see error code for details

• Dependency

  • Header File: mi_sys.h, mi_md.h, mi_od.h, mi_vg.h, mi_vdf.h

  • Library File: libOD_LINUX.a, libMD_LINUX.a, libVG_LINUX.a, libmi_vdf.a

• Note

  • VdfChn cannot be specified repeatedly.

  • VdfChn valid range: 0 <= VdfCh < MI_VDF_CHANNEL_MAX.

  • MI_VDF_CHANNEL_MAX is defined in header file mi_vdf_datatype.h.

• Related APIs

  MI_VDF_DestroyChn

2.2.4. MI_VDF_DestroyChn

• Functions

  Destroy the created video detection channel.

• Syntax

  MI_S32 MI_VDF_DestroyChn(MI_VDF_CHANNEL VdfChn);
  

• Parameters

  Parameter Name	Description	Input/Output
  VdfChn	The video detection channel number that has been created.	Input

• Return Value

  • Zero: Successful

  • Non-zero: Failed, see error code for details

• Dependency

  • Header File: mi_sys.h, mi_md.h, mi_od.h, mi_vg.h, mi_vdf.h

  • Library File: libOD_LINUX.a, libMD_LINUX.a, libVG_LINUX.a, libmi_vdf.a

• Note

  • VdfChn cannot be specified repeatedly.

  • VdfChn valid range: 0 <= VdfCh < MI_VDF_CHANNEL_MAX.

  • MI_VDF_CHANNEL_MAX is defined in header file mi_vdf_datatype.h.

2.2.5. MI_VDF_SetChnAttr

• Functions

  Set the video detection channel attributes.

• Syntax

  MI_S32 MI_VDF_SetChnAttr(MI_VDF_CHANNEL VdfChn, const MI_VDF_ChnAttr_t* pstAttr);
  

• Parameters

  Parameter Name	Description	Input/Output
  VdfChn	The video detection channel number that has been created.	Input
  pstAttr	Source port configuration information data structure pointer.	Input

• Return Value

  • Zero: Successful

  • Non-zero: Failed, see error code for details

• Dependency

  • Header File: mi_sys.h, mi_md.h, mi_od.h, mi_vg.h, mi_vdf.h

  • Library File: libOD_LINUX.a, libMD_LINUX.a, libVG_LINUX.a, libmi_vdf.a

• Note

  • VdfChn must be a video detection channel that has been created.

  • This interface specifies the value of the dynamic attribute of the video detection channel (stMdAttr.stMdDynamicParamsIn / stMdAttr.stOdDynamicParamsIn/ stMdAttr.stVgAttr in MI_VDF_ChnAttr_t).

2.2.6. MI_VDF_GetChnAttr

• Functions

  Get the video detection channel attributes.

• Syntax

  MI_S32 MI_VDF_GetChnAttr(MI_VDF_CHANNEL VdfChn, MI_VDF_ChnAttr_t* pstAttr);
  

• Parameters

  Parameter Name	Description	Input/Output
  VdfChn	The video detection channel number that has been created.	Input
  pstAttr	Used to save the returned video detection channel attribute value	Output

• Return Value

  • Zero: Successful

  • Non-zero: Failed, see error code for details

• Dependency

  • Header File: mi_sys.h, mi_md.h, mi_od.h, mi_vg.h, mi_vdf.h

  • Library File: libOD_LINUX.a, libMD_LINUX.a, libVG_LINUX.a, libmi_vdf.a

2.2.7. MI_VDF_EnableSubWindow

• Functions

  Enable/disable the specified video detection channel sub-window.

• Syntax

  MI_S32 MI_VDF_EnableSubWindow(MI_VDF_CHANNEL VdfChn, MI_U8 u8Col, MI_U8 u8Row, MI_U8 u8Enable);
  

• Parameters

  Parameter Name	Description	Input/Output
  VdfChn	The video detection channel number that has been created.	Input
  u8Col	The row address of the sub window (reserved, not used)	Input
  u8Row	Sub-window column address (reserved, not used)	Input
  u8Enable	Sub-window enable control flag	Input

• Return Value

  • Zero: Successful

  • Non-zero: Failed, see error code for details

• Dependency

  • Header File: Mi_sys.h, mi_md.h, mi_od.h, mi_vg.h, mi_vdf.h

  • Library File: libOD_LINUX.a, libMD_LINUX.a, libVG_LINUX.a, libmi_vdf.a

• Note

  • u8Col and u8Row are reserved for upward compatibility of API interface, and have no practical effect. You can set them to 0 directly.

  • There are currently three working modes supported: MD, OD, and VG.

  • The MI_VDF_Run/MI_VDF_Stop function acts on all video detection channels in the same working mode (MD/OD/VG), and the MI_VDF_EnableSubWindow function acts on a single video detection channel. To run a single VdfChn, call MI_VDF_Run function and MI_VDF_EnableSubWindow (u8Enable = True). Note that when you call MI_VDF_Stop function or MI_VDF_EnableSubWindow (u8Enable = False), the video detection channel will stop.

2.2.8. MI_VDF_Run

• Functions

  Start running video detection working mode.

• Syntax

  MI_S32 MI_VDF_Run(MI_VDF_WorkMode_e enWorkMode);
  

• Parameters

  Parameter Name	Description	Input/Output
  enWorkMode	Video detection (MD/OD/VG) working mode.	Input

• Return Value

  • Zero: Successful

  • Non-zero: Failed, see error code for details

• Dependency

  • Header File: mi_sys.h, mi_md.h, mi_od.h, mi_vg.h, mi_vdf.h

  • Library File: libOD_LINUX.a, libMD_LINUX.a, libVG_LINUX.a, libmi_vdf.a

• Note

  • The work mode only supports MD/OD/VG.

  • MI_VDF_Run/MI_VDF_Stop function acts on all video detection channels in the same working mode (MD/OD/VG).

2.2.9. MI_VDF_Stop

• Functions

  Stop running video detection (MD/OD/VG) working mode.

• Syntax

  MI_S32 MI_VDF_Stop(MI_VDF_WorkMode_e enWorkMode);
  

• Parameters

  Parameter Name	Description	Input/Output
  enWorkMode	Video detection (MD/OD/VG) working mode.	Input

• Return Value

  • Zero: Successful

  • Non-zero: Failed, see error code for details

• Dependency

  • Header File: mi_sys.h, mi_md.h, mi_od.h, mi_vg.h, mi_vdf.h

  • Library File: libOD_LINUX.a, libMD_LINUX.a, libVG_LINUX.a, libmi_vdf.a

• Note

  • The work mode only supports MD/OD/VG.

  • MI_VDF_Run/MI_VDF_Stop function acts on all video detection channels in the same working mode (MD/OD/VG).

2.2.10. MI_VDF_GetResult

• Functions

  Get the detection result of the specified video detection channel.

• Syntax

  MI_S32 MI_VDF_GetResult(MI_VDF_CHANNEL VdfChn, MI_VDF_Result_t* pstVdfResult, MI_S32 s32MilliSec);
  

• Parameters

  Parameter Name	Description	Input/Output
  VdfChn	The video detection channel number that has been created.	Input
  pstVdfResult	Save the returned video detection results	Output
  s32MilliSec	Input time parameter**(reserved, not used currently)**	Input

• Return Value

  • Zero: Successful

  • Non-zero: Failed, see error code for details

• Dependency

  • Header File: mi_sys.h, mi_md.h, mi_od.h, mi_vg.h, mi_vdf.h

  • Library File: libOD_LINUX.a, libMD_LINUX.a, libVG_LINUX.a, libmi_vdf.a

• Note

  • VdfChn must be a video detection channel that has been created.

  • s32MilliSec is reserved for upward compatibility of API interface, and has no practical effect. You can set it to 0 directly.

  • The pointer used to save the returned result cannot be Null.

2.2.11. MI_VDF_PutResult

• Functions

  Release the detection result of the specified video detection channel.

• Syntax

  MI_S32 MI_VDF_PutResult(MI_VDF_CHANNEL VdfChn, MI_VDF_Result_t* pstVdfResult);
  

• Parameters

  Parameter Name	Description	Input/Output
  VdfChn	The video detection channel number that has been created.	Input
  pstVdfResult	Specify the result parameters that need to release the video channel	Input

• Return Value

  • Zero: Successful

  • Non-zero: Failed, see error code for details

• Dependency

  • Header File: mi_sys.h, mi_md.h, mi_od.h, mi_vg.h, mi_vdf.h

  • Library File: libOD_LINUX.a, libMD_LINUX.a, libVG_LINUX.a, libmi_vdf.a

• Note

  VdfChn must be a video detection channel that has been created.

2.2.12. MI_VDF_GetLibVersion

• Functions

  Obtain the MD/OD/VG library version number.

• Syntax

  MI_S32 MI_VDF_GetLibVersion(MI_VDF_CHANNEL VdfChn, MI_U32* u32VDFVersion);
  

• Parameters

  Parameter Name	Description	Input/Output
  VdfChn	The video detection channel number that has been created.	Input
  u32VDFVersion	Pointer for saving the version number (cannot be Null).	Output

• Return Value

  • Zero: Successful

  • Non-zero: Failed, see error code for details

• Dependency

  • Header File: mi_sys.h, mi_md.h, mi_od.h, mi_vg.h, mi_vdf.h

  • Library File: libOD_LINUX.a, libMD_LINUX.a, libVG_LINUX.a, libmi_vdf.a

2.2.13. MI_VDF_GetDebugInfo

• Functions

  Get the MD/OD/VG debug info.

• Syntax

  MI_S32 MI_VDF_GetDebugInfo(MI_VDF_CHANNEL VdfChn, MI_VDF_DebugInfo_t *pstDebugInfo);
  

• Parameters

  Parameter Name	Description	Input/Output
  VdfChn	The video detection channel number that has been created.	Input
  u32VDFVersion	Pointer for saving the VDF debug info (cannot be Null)	Output

• Return Value

  • Zero: Successful

  • Non-zero: Failed, see error code for details

• Dependency

  • Header File: mi_sys.h, mi_md.h, mi_od.h, mi_vg.h, mi_vdf.h

  • Library File: libOD_LINUX.a, libMD_LINUX.a, libVG_LINUX.a, libmi_vdf.a

• Note

  Only supports VG mode.

---

3. DATA TYPE

3.1. VDF Structure Description

Data structure	Definition
MI_VDF_WorkMode_e	An enumerated type that defines the working mode of VDF
MI_VDF_Color_e	Define the enumerated type of the video detection channel input source
MI_VDF_ODWindow_e	Enumerated type of the number of child windows of the screen when defining OD
MI_MD_Result_t	Structure that defines the MD result
MI_OD_Result_t	Structure that defines the OD result
MI_VG_Result_t	Structure that defines the VG result
MI_VDF_Result_t	Structure that defines the corresponding result of the VDF working mode
MI_VDF_MdAttr_t	Structure that defines MD channel attributes
MI_VDF_OdAttr_t	Structure that defines OD channel attributes
MI_VDF_VgAttr_t	Structure that defines VG channel attributes
MI_VDF_ChnAttr_t	Structure that defines the channel's working mode properties
MDRST_STATUS_t	Structure that defines a motion detection result of a detection sub-window region
MI_MD_ResultSize_t	Structure that defines the length of the MD result

3.1.1. MI_VDF_WorkMode_e

• Description

  Define VDF work mode enumeration type.

• Definition

  typedef enum
  
  {
  
      E_MI_VDF_WORK_MODE_MD = 0,
  
      E_MI_VDF_WORK_MODE_OD,
  
      E_MI_VDF_WORK_MODE_VG,
  
      E_MI_VDF_WORK_MODE_MAX
  
  }MI_VDF_WorkMode_e;
  

• Members

  Member Name	Description
  E_MI_VDF_WORK_MODE_MD	MD working mode
  E_MI_VDF_WORK_MODE_OD	OD working mode
  E_MI_VDF_WORK_MODE_VG	VG working mode
  E_MI_VDF_WORK_MODE_MAX	Error code identification of working mode

3.1.2. MI_VDF_Color_e

• Description

  Define the enumerated type of the video detection channel input source

• Definition

  typedef enum
  
  {
  
      E_MI_VDF_COLOR_Y = 1,
  
      E_MI_VDF_COLOR_MAX
  
  } MI_VDF_Color_e;
  

• Members

  Member Name	Description
  E_MI_VDF_COLOR_Y	Correct identification of the input source type of the video detection channel
  E_MI_VDF_COLOR_MAX	Video detection channel input source type error code identification

3.1.3. MI_VDF_ODWindow_e

• Description

  Enumerated type of the number of child windows of the screen when defining OD

• Definition

  typedef enum
  
  {
  
      E_MI_VDF_ODWINDOW_1X1 = 0,
  
      E_MI_VDF_ODWINDOW_2X2,
  
      E_MI_VDF_ODWINDOW_3X3,
  
      E_MI_VDF_ODWINDOW_MAX
  
  } MI_VDF_ODWindow_e;
  

• Members

  Member Name	Description
  E_MI_VDF_ODWINDOW_1X1	The OD screen is divided into 1 sub-window
  E_MI_VDF_ODWINDOW_2X2	The OD screen is divided into 2x2 sub-windows
  E_MI_VDF_ODWINDOW_3X3	The OD screen is divided into 3x3 sub-windows
  E_MI_VDF_ODWINDOW_MAX	Error code identification of the number of OD screen sub-windows

3.1.4. MI_MD_Result_t

• Description

  A structure that defines the MD result.

• Definition

  typedef struct MI_MD_Result_s
  
  {
  
      MI_U64 u64Pts;        //The PTS of Image
  
      MI_U8 u8Enable;       // =1 indicates that the result value is valid
  
      MI_MD_ResultSize_t stSubResultSize;
  
      MDRST_STATUS_t* pstMdResultStatus; //The MD result of Status
  
      MDSAD_DATA_t* pstMdResultSad; //The MD result of SAD
  
      MDOBJ_DATA_t* pstMdResultObj; //The MD result of Obj
  
  }MI_MD_Result_t;
  

• Members

  Member Name	Description
  u64Pts	Image display timestamp
  u8Enable	Indicates whether the channel is enabled
  u8Reading	Indicates that the result is being read by the application layer
  stSubResultSize	Describe the Size of the Sad、Obj and ReasultStauts substructures returned by the MD
  pstMdResultStatus	Describe whether motion is detected in each area of the MD, output in MB, 0-block does not detect motion, and 255-block detects motion
  pstMdResultSad	Describe the Sad value of the MD, For example, if the input image is 640*360 and MB:8x8, (640/8)x(360/8)=80x45 SAD values will be obtained
  pstMdResultObj	Describe the CCL value of MD, and the result of the connectivity operation is in units of objects

3.1.5. MI_OD_Result_t

• Description

  The structure that defines the OD result.

• Definition

  typedef struct MI_OD_Result_s
  
  {
  
      MI_U8 u8Enable;
  
      MI_U8 u8WideDiv;      //The number of divisions of window in horizontal direction
  
      MI_U8 u8HightDiv;     //The number of divisions of window in vertical direction
  
      MI_U8 u8DataLen;      //OD detect result readable size
  
      MI_U64 u64Pts;       //The PTS of Image
  
      MI_S8 u8RgnAlarm[3][3];  //The OD result of the sub-window
  
      MI_S8 s8OdStatus;    //The OD detect status
  
  }MI_OD_Result_t;
  

• Members

  Member name	description
  u8Enable	Indicates whether the channel is enabled
  u8WideDiv	Get the number of windows in the horizontal direction of the OD
  u8HightD iv	Get the number of windows in the vertical direction of the OD
  u8DataLen	Set the readable size of the OD test results
  u64Pts	Image display time
  u8RgnAlarm[3][3]	OD sub window information result 0: The window is not blocked 1: The window is blocked 2: Insufficient window feature -1: Fail
  s8OdStatus	1. s8OdStatus = 0: No od detected 2. s8OdStatus = 1: Od detected 3. s8OdStatus = -1: Run od error

• Note

  If the user needs to obtain the value of s8OdStatus, it can be judged by the status value 0 or 1. When it is -1, it means that OD has failed to run, and MI_VDF_GetResult failed, so that the value of s8OdStatus cannot be obtained.

3.1.6. MI_VG_Result_t

• Description

  he structure that defines the VG result.

• Definition

  typedef struct _MI_VgResult_t
  
  {
  
      int32_t alarm[MAX_NUMBER];
  
      int32_t alarm_cnt;
  
      MI_VgBoundingBox_t bounding_box[20];
  
  } MI_VgResult_t;
  

• Members

  Member Name	Description
  alarm	Describe the test results of VG #define MAX_NUMBER 4
  alarm_cnt	Describe the number of VG alerts
  bounding_box	Describe the object information that triggered the VG alarm

3.1.7. MI_VDF_Result_t

• Description

  A structure that defines the result of the VDF working mode.

• Definition

  typedef struct MI_VDF_Result_s
  
  {
  
      MI_VDF_WorkMode_e enWorkMode;
  
      VDF_RESULT_HANDLE handle;
  
      union
  
      {
  
      ​    MI_MD_Result_t stMdResult;
  
      ​    MI_OD_Result_t stOdResult;
  
      ​    MI_VG_Result_t stVgResult;
  
      };
  
  }MI_VDF_Result_t;
  

• Members

  Member name	description
  enWorkMode	VDF working mode (MD/OD/VG)
  Handle	Save result handle
  stMdResult	Structure describing the MD result
  stOdResult	Structure describing the OD result
  stVgResult	Structure describing the VG result

3.1.8. MI_VDF_MdAttr_t

• Description

  Structure that defines MD channel properties

• Definition

  typedef struct MI_VDF_MdAttr_s
  {
      MI_U8  u8Enable;
      MI_U8  u8MdBufCnt;
      MI_U8  u8VDFIntvl;
      MI_U32 u32RstBufSize;
      MI_MD_ResultSize_t stSubResultSize;
      MDCCL_ctrl_t ccl_ctrl;
      MI_MD_static_param_t  stMdStaticParamsIn;
      MI_MD_param_t stMdDynamicParamsIn;
  }MI_VDF_MdAttr_t;
  

• Members

  Member name	description
  u8Enable	Indicates whether the channel is enabled
  u8MdBufCnt	Set the number of MD results that can be cached MD result cache number range: [1, 8 ] static attribute
  u8VDFIntvl	Detection interval value range: [0, 29] , in frames, dynamic attribute
  u32RstBufSize	The total size of the MD return results
  stSubResultSize	MD returns the result of each substructure (Sad value, CCL value, ResultStatus) size
  Ccl_ctrl	c cl _ctrl property setting
  stMdStaticParamsIn	MD static attribute parameter setting
  stMdDynamicParamsIn	MD dynamic attribute parameter setting

3.1.9. MI_VDF_OdAttr_t

• Description

  Structure that defines OD channel attributes

• Definition

  typedef struct MI_VDF_OdAttr_s
  {
      MI_U8 u8Enable;
      MI_U8 u8OdBufCnt;
      MI_U8 u8VDFIntvl;
      MI_U32 u32RstBufSize;
      MI_OD_static_param_t stOdStaticParamsIn;
      MI_OD_param_t stOdDynamicParamsIn;
  }MI_VDF_OdAttr_t;
  

• Members

  Member name	description
  u8Enable	Indicates whether the channel is enabled
  u8OdBufCnt	OD result cached number range: [1, 16] static attribute
  u8VDFIntvl	Detection interval value range: [0, 29] , in frames, dynamic attribute
  u32RstBufSize	The total size of the OD return results
  stOdDynamicParamsIn	O D dynamic attribute parameter setting
  stOdStaticParamsIn	O D static attribute parameter setting

3.1.10. MI_VDF_VgAttr_t

• Description

  Structure that defines VG channel properties

• Definition

  typedef struct MI_VDF_VgAttr_s
  {
      MI_U8 u8Enable;
      MI_U8 u8VgBufCnt;
      MI_U8 u8VDFIntvl;
      MI_U32 u32RstBufSize;
  
      MI_U16 width;
      MI_U16 height;
      MI_U16 stride;
  
      float object_size_thd;
      uint8_t indoor;
      uint8_t function_state;
      uint16_t line_number;
      MI_VgLine_t line[4];
      MI_VgRegion_t vg_region;
  
      MI_VgSet_t stVgParamsIn;
  } MI_VDF_VgAttr_t;
  

• Members

  Member name	description
  u8Enable	Indicates whether the channel is enabled
  u8VgBufCnt	VG result buffered number range: [1, 8 ] static attribute
  u8VDFIntvl	Detection interval value range: [0, 29] , in frames, dynamic attribute
  u32RstBufSize	The total size of the result returned by the VG
  Width	Image width
  Height	Image height
  Stride	Image of stride
  Object_size_thd	Decided to filter out the percentage of objects in the area of interest (If object_size_thd = 1 indicates that the object area is less than one percent of the region of interest in the image frame, it will be ignored.)
  Indoor	Whether the camera is mounted indoors or outdoors, 1- indoor, 0- outdoor
  Function_state	Set a detection mode for virtual line segments and regional intrusions: 1) VG_VIRTUAL_GATE, indicates that the mode is a virtual line segment 2) VG_REGION_INVASION, indicates that the pattern is a zone intrusion
  Line_number	Set the number of virtual segments, range: [ 1-4 ]
  Line[4]	Show the structure of the virtual line, a maximum of four may be provided
  Vg_region	Related parameters indicating regional invasion
  stVgParamsIn	Vg attribute parameter structure, no need to set ,returned by API

3.1.11. MI_VDF_ChnAttr_t

• Description

  A structure that defines the channel's working mode properties.

• Definition

  typedef struct MI_VDF_ChnAttr_s
  {
      MI_VDF_WorkMode_e enWorkMode;
      union
      {
          MI_VDF_MdAttr_t stMdAttr;
          MI_VDF_OdAttr_t stOdAttr;
          MI_VDF_VgAttr_t stVgAttr;
      };
  

• Members

  Member name	description
  enWorkMode	Working mode (motion detection , occlusion detection , electronic fence) static properties
  stMdAttr	Motion detection attribute
  stOdAttr	Occlusion detection attribute
  stVgAttr	Electronic fence properties

3.1.12. MDRST_STATUS_t

• Description

  A structure defining a motion detection result of a detection sub-window region

• Definition

  typedef struct MDRST_STATUS_s
  
  {
  
      MI_U8 *paddr;
  
  } MDRST_STATUS_t;
  

• Members

  Member Name	Description
  paddr	Directed motion detection state buf, each area occupies 1byte 0- block does not detect motion, 255- block detects motion

3.1.13. MI_MD_ResultSize_t

• Description

  Structure that defines the length of the MD result

• Definition

  typedef struct MI_MD_ResultSize_s
  
  {
  
      MI_U32 u32RstStatusLen;
  
      MI_U32 u32RstSadLen;
  
      MI_U32 u32RstObjLen;
  
  }MI_MD_ResultSize_t;
  

• Members

  Member Name	Description
  u32RstStatusLen	Describe the length of the MD detected motion state value
  u32RstSadLen	Describe the length of the Sad value of MD
  u32RstObjLen	Describe the length of the CCL value of MD

3.3. MD Structure Description

Data Type	Definition
MDMB_MODE_e	Macro block size of enumeration value
MDSAD_OUT_CTRL_e	Enumeration value of the SAD output format
MDALG_MODE_e	The operation mode enumeration value of the CCL connected region can be used for CCL operation according to the foreground result or the SAD result.
MDCCL_ctrl_t	Parameter structure that controls CCL operation
MDPreproc_ctrl_t	Contol Parameter for MI_MD_Preproc running. This feature is not currently supported by VDF
MDblock_info_t	Define the ROI coordinate structure of a block.
MDPoint_t	Coordinate structure
MDROI_t	MD detection area structure
MDSAD_DATA_t	The structure of the MI_MD_ComputeImageSAD function output.
MDOBJ_t	Define the information of the connected area: the area and the coordinate position of the smallest bounding rectangle
MDOBJ_DATA_t	Structure of CCL output
MI_MD_IMG_t	The image source structure of motion detection is divided into physical and virtual memory address pointers.
MI_MD_IMG64_t	The image source structure of motion detection is divided into 64 bit physical memory address and virtual memory address pointer.
MI_MD_static_param_t	MD static parameter setting structure
MI_MD_param_t	MD dynamic parameter setting structure

3.2.1. MDMB_MODE_e

• Description

  The enumerated size of Micro-block.

• Definition

  typedef enum MDMB_MODE_E
  
  {
  
      MDMB_MODE_MB_4x4   = 0x0,
  
      MDMB_MODE_MB_8x8   = 0x1,
  
      MDMB_MODE_MB_16x16  = 0x2,
  
      MDMB_MODE_BUTT
  
  } MDMB_MODE_e;
  

• Members

  Member Name	Description
  MDMB_MODE_MB_4x4	Use 4x4 Micro-block
  MDMB_MODE_MB_8x8	Use 8x8 Micro-block
  MDMB_MODE_MB_16x16	Use 16x16 Micro-block

3.2.2. MDSAD_OUT_CTRL_e

• Description

  The enumeration value of the SAD output format.

• Definition

  typedef enum MDSAD_OUT_CTRL_E
  
  {
  
      MDSAD_OUT_CTRL_16BIT_SAD  = 0x0,
  
      MDSAD_OUT_CTRL_8BIT_SAD  = 0x1,
  
      MDSAD_OUT_CTRL_BUTT
  
  } MDSAD_OUT_CTRL_e;
  

• Members

  Member Name	Description
  MDSAD_OUT_CTRL_16BIT_SAD	16 bit Output
  MDSAD_OUT_CTRL_8BIT_SAD	8 bit Output

3.2.3. MDALG_MODE_e

• Description

  The operation mode enumeration value of the CCL connected region can be CCL calculated according to the foreground result or the SAD result

• Definition

  typedef enum MDALG_MODE_E
  
  {
  
      MDALG_MODE_FG   = 0x0,
  
      MDALG_MODE_SAD   = 0x1,
  
      MDALG_MODE_FRAMEDIFF = 0x2,
  
      MDALG_MODE_BUTT
  
  } MDALG_MODE_e;
  

• Members

  Member Name	Description
  MDALG_MODE_FG	Foreground mode
  MDALG_MODE_SAD	SAD Mode
  MDALG_MODE_FRAMEDIFF	FrameDifference Mode

• Tips

  The foreground mode (FG) is implemented by software, while the SAD and Frame Diff modes are supported by hardware operators, and the operators need to be aligned with the input image's stride and width of 16 pixels, and height of 2 pixels.

3.2.4. MDCCL_ctrl_t

• Description

  Control the parameter structure of the CCL execution.

• Definition

  typedef struct MDCCL_ctrl_s
  
  {
  
      uint16_t u16InitAreaThr;
  
      uint16_t u16Step;
  
  } MDCCL_ctrl_t;
  

• Members

  Member Name	Description
  u16InitAreaThr	Area threshold(This value is not necessarily useful,but it recommended value is 8)
  u16Step	Increased value of each threshold(This value is not necessarily useful,but it recommended value is 2)

3.2.5. MDPreproc_ctrl_t

• Description

  Contol Parameter for MI_MD_Preproc running. This feature is not currently supported by VDF

• Definition

  typedef struct MDPreproc_ctrl_s
  
  {
  
      uint16_t u16Md_rgn_size;
  
      uint16_t u16Align;
  
  } MDPreproc_ctrl_t;
  

• Members

  Member Name	Description
  u16Md_rgn_size	The threshold value of the size range of the moving area. If it is less than this threshold, the area is not calculated as a dynamic area
  u16Align	Align restriction corresponding to CNN model (HC/HD:32ss align, FD:64align)

3.2.6. MDblock_info_t

• Description

  Define the ROI coordinate structure of a block. This feature is not currently supported by VDF

• Definition

  typedef struct MDblock_info_s
  
  {
  
      uint16_t st_x;
  
      uint16_t st_y;
  
      uint16_t end_x;
  
      uint16_t end_y;
  
  } MDblock_info_t;
  

• Members

  Member Name	Description
  st_x	Upper left X coordinate of ROI
  st_y	Upper left y coordinate of ROI
  end_x	Lower right X coordinate of ROI
  end_y	Lower right y coordinate of ROI

3.2.7. MDPoint_t

• Description

  Coordinate structure.

• Definition

  typedef struct MDPoint_s
  
  {
  
      uint16_t x;
  
      uint16_t y;
  
  } MDPoint_t;
  

• Members

  Member Name	Description
  x	X Coordinate
  y	Y Coordinate

3.2.8. MDROI_t

• Description

  MD detection area structure.

• Definition

  typedef struct MDROI_s
  
  {
  
      uint8_t num;
  
      MDPoint_t pnt[8];
  
  } MDROI_t;
  

• Members

  Member Name	Description
  num	MD detection area points, currently only supported to set to 4 points
  pnt[8]	Four-point coordinates (must be set to rectangle)

• Note

  It is required to be set to a rectangle and the four points have a preceding and backward constraint, num=4, starting from the upper left corner point, and setting the four points clockwise, as shown in the following figure

  

3.2.9. MDSAD_DATA_t

• Description

  The structure of the MI_MD_ComputeImageSAD function output.

• Definition

  typedef struct MDSAD_DATA_s
  
  {
  
      void *paddr;
  
      uint32_t stride;
  
      MDSAD_OUT_CTRL_e enOutCtrl;
  
  } MDSAD_DATA_t;
  

• Members

  Member Name	Description
  paddr	Memory address pointer for storing SAD results
  stride	Image stride
  enOutCtrl	Enumeration value of the SAD output format

3.2.10. MDOBJ_t

• Description

  Define the information of the connected area: the area and the coordinate position of the smallest bounding rectangle.

• Definition

  typedef struct MDOBJ_s
  
  {
  
      uint32_t u32Area;
  
      uint16_t u16Left;
  
      uint16_t u16Right;
  
      uint16_t u16Top;
  
      uint16_t u16Bottom;
  
  } MDOBJ_t;
  

• Members

  Member name	description
  u32Area	Total number of pixels in a single connected area
  u16Left	The top left x coordinate of the smallest rectangle
  u16Right	The lower right x coordinate of the smallest rectangle
  u16Top	The upper left y coordinate of the smallest rectangle
  u16Bottom	The upper left y coordinate of the smallest rectangle

• Note

  For example: {3, 1, 2, 2, 3}, element 1 means that there are three pixels in a connected label, assuming that they are: (1,3), (2,2), (2,3). Element 2 represents the smallest value of x from these pixels, element 3 represents the maximum value of x from these pixels, and elements 4 and 5 take the smallest and maximum value of y from these pixels, respectively.

3.2.11. MDOBJ_DATA_t

• Description

  The structure of the MI_MD_CCL output.

• Definition

  typedef struct MDOBJ_DATA_s
  
  {
  
      uint8_t u8RegionNum;
  
      MDOBJ_t *astRegion;
  
      uint8_t indexofmaxobj;
  
      uint32_t areaofmaxobj;
  
      uint32_t areaoftotalobj;
  
  } MDOBJ_DATA_t;
  

• Members

  Member name	description
  u8RegionNum	Number of connected areas.
  astRegion	Information about the connected area: the area and the coordinate position of the smallest bounding rectangle. The maximum allowable number is 255.
  Indexofmaxobj	Maximum area connected area index value.
  Areaofmaxobj	Maximum area of connected area.
  Areaoftotalobj	The area sum-up of all connected areas.

3.2.12. MI_MD_IMG_t

• Description

  The image source structure of motion detection is divided into physical and virtual memory address pointers.

• Definition

  typedef struct MI_MD_IMG_s
  
  {
  
      void *pu32PhyAddr;
  
      uint8_t *pu8VirAddr;
  
  } MI_MD_IMG_t;
  

• Members

  Member Name	Description
  pu32PhyAddr	Physical memory address pointer, default is NULL
  pu8VirAddr	Virtual memory address pointer

3.2.13. MI_MD_IMG64_t

• Description

  The image source structure of motion detection is divided into 64 bit physical memory address and virtual memory address pointer.

• Definition

  typedef struct MI_MD_IMG64_s
  
  {
  
      uint64_t u64PhyAddr;
  
      uint8_t *pu8VirAddr;
  
  } MI_MD_IMG64_t;
  

• Members

  Member Name	Description
  u64PhyAddr	Physical memory address
  pu8VirAddr	Virtual memory address pointer

3.2.14. MI_MD_static_param_t

• Description

  MD static parameter setting structure.

• Definition

  typedef struct MI_MD_static_param_s
  {
      uint16_t width;
      uint16_t height;
      uint8_t color;
      uint32_t stride;
      MDMB_MODE_e mb_size;
      MDSAD_OUT_CTRL_e sad_out_ctrl;
      MDROI_t roi_md;
      MDALG_MODE_e md_alg_mode;
  } MI_MD_static_param_t;
  

• Members

  Member name	description
  Width	Input image width
  Height	Input image is high
  Stride	Input image of stride
  Color	MD input image type. At present, MD only eats the Y portion, which is fixed to 1 and does not change
  mb_size	Macro block size enumeration value
  Sad_out_ctrl	Enumeration value of the SAD output format
  Roi_md	MD detection area structure
  Md_alg_mode	Operation mode enumeration value of CCL connected area

3.2.15. MI_MD_param_t

• Description

  MD dynamic parameter setting structure.

• Definition

  typedef struct MI_MD_param_s
  
  {
  
      uint8_t sensitivity;
  
      uint16_t learn_rate;
  
      uint32_t md_thr;
  
      uint32_t obj_num_max;
  
      uint8_t LSD_open;
  
  } MI_MD_param_t;
  

• Members

  Member name	description
  Sensitivity	Algorithm sensitivity, range [10, 20, 30, ..... 100] , the larger the value, the more sensitive, the sensitivity of the input is not A multiple of 10 , when the feedback is calculated , there may be a value that is not originally entered , there will be +-1 deviation. Only for VG mode.
  Learn_rate	There are different setting standards for different modes: MDALG_MODE_FG : Unit is millisecond, range [1000-30000], used to control how long the front-end object stops moving before it is used as the background picture. MDALG_MODE_SAD : Range [1, 255], background update weight ratio, recommended setting value: 128. In FG, the time is described, and in the SAD, the weight is described, and the fast learning speed or heavy weight is easy to quickly disarm the alarm. For example, if an object enters the screen and is stationary, if it has a lot of weight or a fast learning speed, the object will be quickly learned to become the background, and then the alarm will be released. On the contrary, it will take a long time to release the alarm.
  Md_thr	When calling MI_MD_CCL connected component mark operation, as the threshold value to determine the effective movement, there are different setting standards for different modes: MDALG_MODE_FG : Set to percentage (%), range [0, 99], judge whether MB alarm threshold value, ex. if MB size 8 * 8, MD_ Thr = 50, if more than 32 pixels in MB are foreground, the MB is judged as valid. MDALG_MODE_SAD : If MB size 8 * 8, MD_ Thr = 50, when the average pixel difference in MB exceeds 50, the MB is judged as valid.
  Obj_num_max	CCL connected area number limit value. Not currently working
  LSD_open	Decide whether to use MI_MD_LightSwitchDetect. The range is [0, 1]. If the LightSwitchDetect() function is called when LSD_open=0, it has no effect. Currently, VDF does not support this function

3.3. OD Structure Description

Enumeration
MI_OD_WIN_STATE	OD detection window results
ODColor_e	Type of OD data source input
ODWindow_e	Type of OD detection window
Structure
ODPoint_t	Coordinate structure
ODROI_t	OD detection area structure
MI_OD_IMG_t	The source structure of the occlusion detection image is divided into physical and virtual memory address pointers.
MI_OD_IMG64_t	The source structure of the occlusion detection image is divided into 64 bit physical memory address and virtual memory address pointer.
MI_OD_static_param_t	OD static parameter setting structure
MI_OD_param_t	OD dynamic parameter setting structure

3.3.1. MI_OD_WIN_STATE

• Description

  The result of the OD detection window.

• Definition

  typedef enum _MI_OD_WIN_STATE
  
  {
  
      MI_OD_WIN_STATE_NON_TAMPER = 0,
  
      MI_OD_WIN_STATE_TAMPER = 1,
  
      MI_OD_WIN_STATE_NO_FEATURE = 2,
  
      MI_OD_WIN_STATE_FAIL = -1,
  
  } MI_OD_WIN_STATE;
  

• Members

  Member name	description
  MI_OD_WIN_STATE_NON_TAMPER	Window is not obscured
  MI_OD_WIN_STATE_TAMPER	Window is occluded
  MI_OD_WIN_STATE_NO_FEATURE	Insufficient window features
  MI_OD_WIN_STATE_FAIL	failure

3.3.2. ODColor_e

• Description

  The type of OD data source input.

• Definition

  typedef enum
  
  {
  
      OD_Y = 1,
  
      OD_COLOR_MAX
  
  } ODColor_e;
  

• Members

  Member Name	Description
  OD_Y	y component in the YUV data source
  OD_COLOR_MAX	Enter the maximum value of the image type

3.3.3. ODWindow_e

• Description

  The type of OD detection window, the recommended value is OD_WINDOW_3X3 for testing.

• Definition

  typedef enum
  
  {
  
      OD_WINDOW_1X1 = 0,
  
      OD_WINDOW_2X2,
  
      OD_WINDOW_3X3,
  
      OD_WINDOW_MAX
  
  } ODWindow_e;
  

• Members

  Member Name	Description
  OD_WINDOW_1X1	1 window
  OD_WINDOW_2X2	4 windows
  OD_WINDOW_3X3	9 windows
  OD_WINDOW_MAX	Maximum window type

3.3.4. ODPoint_t

• Description

  Coordinate structure.

• Definition

  typedef struct ODPoint_s
  
  {
  
      uint16_t x;
  
      uint16_t y;
  
  } ODPoint_t;
  

• Members

  Member Name	Description
  x	X coordinate
  y	Y coordinate

3.3.5. ODROI_t

• Description

  OD detection area structure.

• Definition

  typedef struct ODROI_s
  
  {
  
      uint8_t num;
  
      ODPoint_t pnt[8];
  
  } ODROI_t;
  

• Members

  Member Name	Description
  num	OD detection area points, currently only supported to set to 4 points
  pnt[8]	Four-point coordinates (must be set to rectangle)

• Note

  The requirement is set to a rectangle, num=4 , and the coordinates of the upper left corner are set clockwise in sequence, as shown in the figure.

  

3.3.6. MI_OD_IMG_t

• Description

  The source structure of the occlusion detection image is divided into physical and virtual memory address pointers.

• Definition

  typedef struct MI_OD_IMG_s
  
  {
  
      void *pu32PhyAddr;
  
      uint8_t *pu8VirAddr;
  
  } MI_OD_IMG_t;
  

• Members

  Member Name	Description
  pu32PhyAddr	Entity memory address pointer
  pu8VirAddr	Virtual memory address pointer

3.3.7. MI_OD_IMG64_t

• Description

  The source structure of the occlusion detection image is divided into 64 bit physical memory address and virtual memory address pointer.

• Definition

  typedef struct MI_OD_IMG64_s
  {
      uint64_t u64PhyAddr;
      uint8_t *pu8VirAddr;
  } MI_OD_IMG64_t;
  

• Members

  Member Name	Description
  u64PhyAddr	Entity memory address
  pu8VirAddr	Virtual memory address pointer

3.3.8. MI_OD_static_param_t

• Description

  OD static parameter setting structure.

• Definition

  typedef struct MI_OD_static_param_s
  {
      uint16_t inImgW;
      uint16_t inImgH;
      uint32_t inImgStride;
      ODColor_e nClrType;
      ODWindow_e div;
      ODROI_t roi_od;
      int32_t alpha;
      int32_t M;
      int32_t MotionSensitivity;
  } MI_OD_static_param_t;
  

• Members

  Member name	description
  inImgW	Input image width
  inImgH	Input image is high
  inImgStride	Input image stride
  nClrType	OD input image type
  Div	Type of OD detection window
  Roi_od	OD detection area structure
  Alpha	Controlling the learning rate of the reference image
  M	How many images are updated once for reference images
  MotionSensitivity	Mobile sensitivity setting

• Note

  • Setting range Alpha : 0~10 , it is recommended to set 2 , it is not recommended to change. The alpha coefficient is the image blending coefficient used in the accumulation process of background feature frames, alpha only works when SW is taken, and the hardware operator is internally fixed.

  • Setting range MotionSensitivity: 0~5 , set 5 to be sensitive to slight sway, easy to report; set 0 means better tolerance for slight sway, will not report, this side refers to the slight sway is wind swaying The recommended initial setting is 5 .

  • M is recommended to set 120 , it is not recommended to change. That is, the number of frames in which the background is blended. Before the M frame is reached, the accumulation of the current existing frame number is used to fuse into the background feature, and after the M frame is reached, the new frame replaces the frame that came in earlier, for example, in the comparison of frame 122, the background fusion frame is frame 2~frame 121. And so on

• Tips

  OD has hardware operator support, but the operator needs to be aligned with the input image's stride and width of 16 pixels, and height of 2 pixels.

3.3.9. MI_OD_param_t

• Description

  OD dynamic parameter setting structure.

• Definition

  typedef struct MI_OD_param_s
  
  {
  
      int32_t thd_tamper;
  
      int32_t tamper_blk_thd;
  
      int32_t min_duration;
  
  } MI_OD_param_t;
  

• Members

  Member name	description
  Thd_tamper	Image difference proportional threshold
  Tamper_blk_thd	Image occlusion area number threshold value
  Min_duration	Image difference duration threshold,The unit is frame cnt

• Note

  • Set the range thd_tamper: 0~10. If thd_tamper=3, it means that more than 70% of the picture is occluded.

  • Set the range tamper_blk_thd: MI_OD_Init type parameter corresponds to the window. If it is OD_WINDOW_3X3, the maximum number of tamper_blk_thd cannot exceed 9, that is, the range is 1~9.

  • MI_OD_Init parameters such as window type is OD_WINDOW_3X3 (9 subareas) tamper_blk_thd value 4, when the number of sub-areas is blocked up to 4 MI_OD_Runtriggered only returns a value of 1.

  • The larger the min_duration value, the longer it takes to detect occlusion.

  • The sensitivity of MI_OD_Run can be adjusted by setting tamper_blk_thd and min_duration. The recommended values for high, medium and low are as follows:

    parameter name	high	in	low
    Tamper_blk_thd	2	4	8
    Min_duration	5	15	30

3.4. VG Structure Description

Enumeration
VgFunction	Detect mode enumeration value
VgRegion_Dir	Enumeration value of the direction of the zone intrusion
VgSize_Sensitively	Sensitivity parameter structure
VgDirection_State	Enumerated values of detection rules for regional intrusion
Structure
MI_VG_Point_t	Coordinate point corresponding structure
MI_VgLine_t	Structure describing virtual segments and directions
MI_VgRegion_t	Describe the structure of setting up zone intrusion
MI_VgSet_t	Vg corresponds to the structure of the parameter settings
MI_VgResult_t	Corresponding structure of Vg detection result
MI_VgBoundingBox_t	Object size structure corresponding to Vg test results
MI_VgDetectThd	Vg establish background information structure parameter setting

3.4.1. VgFunction

• Description

  The enumeration value of the detection mode.

• Definition

  typedef enum _VgFunction
  
  {
  
      VG_VIRTUAL_GATE   = 2,
  
      VG_REGION_INVASION  = 3
  
  } VgFunction;
  

• Members

  Member name	description
  VG_VIRTUAL_GATE	Indicates that the mode is a virtual line segment
  VG_REGION_INVASION	Indicates that the pattern is a regional invasion

3.4.2. VgRegion_Dir

• Description

  The enumeration value of the direction of the zone intrusion.

• Definition

  typedef enum _VgRegion_Dir
  
  {
  
      VG_REGION_ENTER   = 0,
  
      VG_REGION_LEAVING  = 1,
  
      VG_REGION_CROSS   = 2
  
  } VgRegion_Dir;
  

• Members

  Member name	description
  VG_REGION_ENTER	Indicates that an alarm is triggered before entering the alarm zone
  VG_REGION_LEAVING	Indicates that you want to leave the alert area to trigger an alert
  VG_REGION_CROSS	Indicates that an alarm is triggered as soon as it crosses the alarm zone

3.4.3. VgSize_Sensitively

• Description

  Sensitivity parameter structure.

• Definition

  typedef enum _VgSize_Sensitively
  
  {
  
      VG_SENSITIVELY_MIN    = 0,
  
      VG_SENSITIVELY_LOW   = 1,
  
      VG_SENSITIVELY_MIDDLE  = 2,
  
      VG_SENSITIVELY_HIGH   = 3,
  
      VG_SENSITIVELY_MAX   = 4
  
  } VgSize_Sensitively;
  

• Member

  Member name	Description
  VG_SENSITIVELY_MIN	Detect objects larger than 30%
  VG_SENSITIVELY_LOW	Detect objects larger than 10%
  VG_SENSITIVELY_MIDDLE	Detect objects larger than 5%
  VG_SENSITIVELY_HIGH	Detect objects larger than 1%
  VG_SENSITIVELY_MAX	Detect objects larger than 0.5%

3.4.4. VgDirection_State

• Description

  Enumerated values of detection rules for regional intrusion.

• Definition

  typedef enum _VgDirection_State
  
  {
  
      VG_SPEC_DIRECTION_CLOSE  = 0,
  
      VG_SPEC_DIRECTION_OPEN  = 1
  
  } VgDirection_State;
  

• Member

  Member name	Description
  VG_SPEC_DIRECTION_CLOSE	The closed result will distinguish the function of entering and leaving (Under the premise of setting VG_REGION_CROSS, only the alarm is detected, and there is no function of judging the direction.)
  VG_SPEC_DIRECTION_OPEN	The opening result will distinguish the function of entering and leaving (under the premise of setting VG_REGION_CROSS, it can be judged whether the alarm direction is entering or leaving.)

3.4.5. MI_VG_Point_t

• Description

  Describe the structure of the virtual line segment and direction.

• Definition

  typedef struct _VG_Point_t
  
  {
  
      int32_t x;
  
      int32_t y;
  
  } MI_VG_Point_t;
  

• Members

  Member Name	Description
  x	X Coordinate
  y	Y Coordinate

3.4.6. MI_VgLine_t

• Description

  Describe the structure of the virtual line segment and direction.

• Definition

  typedef struct _VG_Line_t
  
  {
  
      MI_VG_Point_t px;  //point x
  
      MI_VG_Point_t py;  //point y
  
      MI_VG_Point_t pdx; //point direction x
  
      MI_VG_Point_t pdy; //point direction y
  
  } MI_VgLine_t;
  

• Members

  Member name	description
  Px	First line point
  Py	Second line point
  Pdx	First direction point
  Pdy	Second direction point

• Note

  • It is built with line segment points and direction points, and the alarm is only generated when it passes through a specific direction, or when it is crossed in any direction. The condition of the alarm is that the center of the object passes through.

  • Specific direction: The first direction point to the second direction point, and the alarm will be generated only when the specified direction is crossed.

  • Any Direction: Set both direction points to the same side of the line segment, or the two direction points set the same coordinates, or the two coordinate points and the line segment are on the same line.

  • In the virtual segment mode, you can set up to four line segments, each corresponding to two direction points, in the same way as above.

3.4.7. MI_VgRegion_t

• Description

  Describe the structure of setting up zone intrusion.

• Definition

  typedef struct _VG_Region_t
  {
      MI_VG_Point_t p_one;    //point one
      MI_VG_Point_t p_two;    //point two
      MI_VG_Point_t p_three;  //point three
      MI_VG_Point_t p_four;   //point four
  
      int32_t region_dir;       //Region direction;
      int32_t spec_dir_state;
      int32_t point_num;
      MI_VG_Point_t reg_coor[MAX_NUMBER];  //Region coordinate
  } MI_VgRegion_t;
  

• Members

  Member name	description
  P_one	Describe the first point of the area
  P_two	Describe the second point of the area
  P_three	Describe the third point of the area
  P_four	The fourth point of the description area
  Region_dir	Set the direction of the area intrusion
  spec_dir_state	Set whether to distinguish between entering and leaving when region_dir = VG_REGION_CROSS
  point_num	Set the point number of polygon region
  reg_coor[MAX_NUMBER]	Set the coordinate of the point of polygon region, wherein MAX_NUMBER is 10

3.4.8. MI_VgSet_t

• Description

  Vg corresponds to the structure of the parameter settings.

• Definition

  typedef struct _MI_VgSet_t
  {
      //Common Information
      float object_size_thd;
      uint16_t line_number;
      uint8_t indoor;
  
      //Line info
      MI_VG_Point_t fp[MAX_NUMBER];   //First point
      MI_VG_Point_t sp[MAX_NUMBER];   //Second point
      MI_VG_Point_t fdp[MAX_NUMBER];  //First direction point
      MI_VG_Point_t sdp[MAX_NUMBER];  //Second direction point
  
      //Function
      uint8_t function_state;
  
      //Region info
      MI_VG_Point_t first_p;   //First point
      MI_VG_Point_t second_p;  //Second point
      MI_VG_Point_t third_p;   //Third point
      MI_VG_Point_t fourth_p;  //Fourth point
  
      //Region direction
      uint8_t region_direction;
  
      //Magic_number
      int32_t magic_number;
  
      int32_t region_spdir_state;
  } MI_VgSet_t;
  

• Members

  Member name	description
  Object_size_thd	Decided to filter out objects of interest to the area accounted for a percentage threshold
  Line_number	Set the number of virtual segments, range: [1-10]
  Indoor	The camera is installed indoors or outdoors, 1- indoor, 0- outdoor
  Fp[MAX_NUMBER]	The first point of the line array
  Sp[MAX_NUMBER]	Second line point array
  Fdp[MAX_NUMBER]	Third line segment point array
  Sdp[MAX_NUMBER]	Fourth line segment point array
  Function_state	Set virtual line segment and area intrusion detection mode
  First_p	The first point in the invading area
  Second_p	Second point in the invading area
  Third_p	The third point of the invading area
  Fourth_p	Invasion area is four points lower
  Region_direction	Related parameters indicating regional invasion
  Magic_number	Magic Number, The internal debug parameters of the algorithm do not require software configuration
  region_spdir_state	The relevant rule direction parameters of the invading area

3.4.9. MI_VgResult_t

• Description

  The structure corresponding to the Vg detection result.

• Definition

  typedef struct _MI_VgResult_t
  {
      int32_t alarm[MAX_NUMBER];
      int32_t alarm_cnt;
      MI_VgBoundingBox_t bounding_box[20];
  } MI_VgResult_t;
  

• Members

  Member Name	Description
  alarm[MAX_NUMBER]	Vg alarm result, MAX_NUMBER represents the number of edges, 1: the edge triggers an alarm, and 0: the edge does not trigger an alarm
  alarm_cnt	Vg alarm times, number of edges on which the alarm was triggered
  bounding_box	The object information for Vg trigger an alarm, Each edge corresponds to the size of the object that triggers the alarm

3.4.10. MI_VgBoundingBox_t

• Description

  Object size structure corresponding to Vg test results

• Definition

  typedef struct _MI_VgBoundingBox_t
  
  {
  
      int32_t up;
  
      int32_t down;
  
      int32_t left;
  
      int32_t right;
  
  } MI_VgBoundingBox_t;
  

• Member

  Member name	Description
  up	The minimum y-coordinate information of the object
  down	The maximum y-coordinate information of the object
  left	The minimum x-coordinate information of the object
  right	The maximum x-coordinate information of the object

3.4.11. MI_VgDetectThd

• Description

  Vg establish background information structure parameter setting

• Definition

  typedef struct _MI_VdDetectThd_t
  
  {
  
      uint8_t function_switch;
  
      uint8_t detect_thd;
  
  } MI_VgDetectThd;
  

• Member

  Member name	Description
  function_switch	Custom threshold to establish the functional status of the background (0-off, 1-on)
  detect_thd	Custom threshold, range: [1-50]

---

4. ERROR CODE

Table 4‑1: MI_VDF API Error Codes

Error Code	Macro Definition	Description
0xA0012001	MI_ERR_VDF_INVALID_DEVID	Device ID out of legal range
0xA0012002	MI_ERR_VDF_INVALID_CHNID	Channel group number error or invalid area handle
0xA0012003	MI_ERR_VDF_ILLEGAL_PARAM	Argument out of legal range
0xA0012004	MI_ERR_VDF_EXIST	Repeat a device, channel, or resource that already exists in Create
0xA0012005	MI_ERR_VDF_UNEXIST	Trying to use a setting that does not exist in Destroyed
0xA0012006	MI_ERR_VDF_NULL_PTR	There are empty pointers in the function arguments
0xA0012007	MI_ERR_VDF_NOT_CONFIG	The module is not configured
0xA0012008	MI_ERR_VDF_NOT_SUPPORT	Unsupported parameters or function
0xA0012009	MI_ERR_VDF_NOT_PERM	Operation is not allowed, e.g. an attempt to modify static configuration parameters
0xA001200C	MI_ERR_VDF_NOMEM	Failed to allocate memory, e.g. insufficient system memory
0xA001200D	MI_ERR_VDF_NOBUF	Allocation cache failed, e.g. the requested data buffer is too large
0xA001200E	MI_ERR_VDF_BUF_EMPTY	No data in the buffer
0xA001200F	MI_ERR_VDF_BUF_FULL	Full data in buffer
0xA0012010	MI_ERR_VDF_NOTREADY	The system did not initialize or load the module
0xA0012011	MI_ERR_VDF_BADADDR	The address is illegal
0xA0012012	MI_ERR_VDF_BUSY	The system is busy
0xA0012013	MI_ERR_VDF_CHN_NOT_STARTED	The channel did not start.
0xA0012014	MI_ERR_VDF_CHN_NOT_STOPED	The channel has not stopped.
0xA0012015	MI_ERR_VDF_NOT_INIT	The module is not initialized
0xA0012016	MI_ERR_VDF_INITED	Module has been initialized
0xA0012017	MI_ERR_VDF_NOT_ENABLE	Channel or port is not enabled
0xA0012018	MI_ERR_VDF_NOT_DISABLE	Channel or port is not disabled
0xA0012019	MI_ERR_VDF_TIMEOUT	Timeout
0xA001201A	MI_ERR_VDF_DEV_NOT_STARTED	The device did not start
0xA001201B	MI_ERR_VDF_DEV_NOT_STOPED	The device is not stopped
0xA001201C	MI_ERR_VDF_CHN_NO_CONTENT	There is no information in the channel
0xA001201D	MI_ERR_VDF_NOVASAPCE	Failed to map virtual address
0xA001201E	MI_ERR_VDF_NOITEM	There is no record
0xA001201F	MI_ERR_VDF_FAILED	Unknown error

---

5. PROCFS INTRODUCTION

5.1. cat

• Debug info

  # cat /proc/mi_modules/mi_vdf/mi_vdf0
  

• Debug info analysis

  Record the current status of vdf and related attributes, which can be obtained dynamically for debugging and testing.

• Parameter Description

  

  Table 1-1:

  Parameter		Description
  Common info for mi_vdf	ChnNum	The maximum channel number of the device
  	EnChnNum	The enabled channel number of the device
  	PassNum	The pass number of the device
  	InPortNum	The inport number of the device
  	OutPortNum	The outport number of the device
  	CollectSize	The number of allocators contained in the AllocatorCollection corresponding to the device
  Common info for mi_vdf only dump enabled chn	ChnId	Channel id
  	PassNum	Pass number
  	EnInPNum	The enabled input port number of the channel
  	EnOutPNum	The enabled output port number of the channel
  	MMAHeapName	If there is SetChnMMAConf, the value is the corresponding mma heap name; if it is not set, the value is NULL
  	current_buf_size	The size of the currently requested memory
  	Peak_buf_size	Peak memory usage
  	user_pid	Current process pid

  

  Table 1-2:

  Parameter		Description
  Input port bind info (only dump enabled Input port)	ChnId	The channel id of the input port
  	PassId	The pass id of the device
  	PortId	The InputPort id of the device
  	user_buf_quota	The quota buff number of the InputPort
  	UsrInjectQ_cnt	The buff number in UsrInjectBufQueue of the InputPort
  	BindInQ_cnt	The buff number in BindInputBufQueue of the InputPort
  	TotalPendingBuf_size	The buff total size in cur_working_input_queue of the InputPort
  	usrLockedInjectCnt	The buff count currently taken by the user layer
  	newPulseQ_cnt	The buff number in new_pulse_fifo_inputqueue of the InputPort
  	nextTodoPulseQ_cnt	The buff number in next_todo_pulse_inputqueue of the InputPort
  	curWorkingQ_cnt	The buff number in cur_working_input_queue of the InputPort
  	workingTask_cnt	The buff number in input_working_tasklist of the InputPort
  	lazzyRewindTask_cnt	The buff number in lazzy_rewind_inputtask_list of the InputPort (The task count that need to be retryed)
  	Enable	1: Enable 0: Disable
  	bind_module_id	The module id of the output port binded with this input port
  	bind_module_name	The module name of the output port binded with this input port (For example, the mod name here is divp, divp→vdf)
  	bind_ChnId	The channel id of the output port binded with this input port
  	bind_PortId	The id of the output port binded with this input port
  	bind_Type	0x00000001: Frame (Frame mode, default work mode) 0x00000002: sw_low_latency (Low-latency work mode) 0x00000004: realtime mode (Hardware direct connection work mode) 0x00000008: hw autosync (Before and after handshake, the buffer size is consistent with the image resolution) 0x00000010: hw ring (Before and after handshake, ring buffer depth can be adjusted)
  	bind_Param	Binded parameter
  	enable	Enable
  	SrcFrmrate	Source frame rate
  	DstFrmrate	Target frame rate
  	GetFrame/Ms	Actual frame rate/Time used
  	FPS	Actual frame rate
  	FinishCnt	Finished count
  	RewindCnt	The task count that need to be retryed

  

  Table 1-3:

  Parameter		Description
  output port bind info(only dump enabled output port)	ChnId	The channel id of the output port
  	PassId	The pass id of the device
  	PortId	The output port of the device
  	usrDepth	The maximum number of output port buf that the user can get
  	BufCntQuota	The quota buff count of the OutputPort
  	usrLockedCnt	The buffer count actually taken by the user layer from UsrGetFifoBufQueue
  	totalOutPortInUsed	The buffer count actually applied for by Output
  	DrvBkRefFifoQ_cnt	The buffer number in DrvBkRefFifoQueue of the OutputPort (For special driver which need to refer back pre-processed buffer)
  	DrvBkRefFifoQ_size	The buffer size occupied in the DrvBkRefFifoQueue of the OutputPort
  	UsrGetFifoQ_cnt	The buffer count dumped in the UsrGetFifoBufQueue of the OutputPort
  	UsrGetFifoQ_size	The buffer size dumped in the UsrGetFifoBufQueue of the OutputPort
  	UsrGetFifoQ_seqnum	The buffer count got by the OutputPort
  	UsrGetFifoQ_discardnum	The cumulative number of discarded original buffers that the OutputPort discarded due to new buffer requirements.
  	finishedTask_cnt	The task count finished by the OutputPort
  	GetFrame/Ms	Actual frame rate/time used
  	FPS	Actual frame rate
  	FinishCnt	Finished count
  	RewindCnt	The task count that need to be retryed
  	GetTotalCnt	The buffer count got by the OutputPort
  	GetOkCnt	The frame count of E_MI_FRC_OBTAIN

  

  

  Table 1-4:

  Parameter		Description
  DEV info for VDF	VDF Verison	Version information
  	bInited	The Vdf module is initialized or not. (0-Disable, 1-Enable)
  	MDEnable	Enable MD or not. (0-Disable, 1-Enable)
  	ODEnable	Enable OD or not. (0-Disable, 1-Enable)
  	VGEnable	Enable VG or not. (0-Disable, 1-Enable)
  CHN common info for VDF (only dump created chn)	ChnId	Channel id
  	eStatus	0: Destroy 1: created 2: start 3: stop
  	phandle	OD/MD/VG init handle pointer
  	FrameCnt	Frame count (Not used currently)
  	FrameInv	Interval frame (Not used currently)
  	YImgSize	A yImage size(height*width) size sent from the previous stage
  Calc info for VDF (only dump created chn)	ChnId	Channel Id
  	CalcGetCnt	The number of input image obtained successfully by vdf
  	CalcGetFailCnt	The number of input image obtained unsuccessfully by vdf
  	CalcDropCnt	The number of input image discarded by vdf
  	CalcDoneCnt	The number of input image processed successfully by vdf
  	Fps	The frame rate of the input image /td>
  	AlgoTime	The time consumed by the MD/OD/VG algorithm to process a frame(ms)
  Rst info for VDF (only dump created chn )	ChnId	Channel Id
  	RstGetCnt	The number of times the user succeeded in obtaining the algorithm result
  	RstGetFailCnt	The number of times the user failed to obtain the algorithm result
  	RstDropCnt	The number of times the algorithm result was discarded (discarded by new data)
  	RstPutCnt	The number of times the user released the memory used by the algorithm result
  MD Attr info for VDF (only dump created chn)	ChnId	Channel id
  	[c_param] Common parameter	Enable	Enable current channel or not
  		MdBufCnt	The Buffer count that MD stores the result
  		VDFIntvl	Not used currently
  		RstBufSize	The size of the result returned at one time
  		SubRstObjLen	The size of the CCL output result
  		SubRstSadLen	The size of the SAD output result
  		SubRstStsLen	The size of the checked result
  		ccl_InitArea	Threshold value of CCL area area
  		ccl_Step	The increase value of the threshold value for each increase in CCL
  	[s_param] Static parameter	width	The Image width input by MD
  		height	The Image height input by MD
  		stride	The Image stride input by MD
  		color	The Image color input by MD
  		mb_size	MD macroblock size. 0: MB_4x4, 1: MB_8x8, 2: MB_16x16
  		sad_out_ctrl	MD SAD output format. 0: 16BIT_SAD, 1: 8BIT_SAD
  		md_alg_mode	Algorithm mode of CCL Unicom area. 0: FG mode, 1: SAD mode
  		roi_md_num	Number of boundary points of detection area
  		Pnt[i].x	The x coordinate of each boundary point of the detection area
  		Pnt[i].y	The y coordinate of each boundary point of the detection area
  	[d_param] Dynamic parameter	sensitivity	Algorithm sensitivity, range [10,20,30,…..100], the larger the value, the more sensitive
  		learn_rate	Unit milliseconds, range [1000,30000], used to control how long the front object stops moving before it is used as the background picture
  		md_thr	The threshold value for judging the movement, there are different setting standards for different modes
  		obj_num_max	CCL limit value for the number of connected regions
  OD Attr info for VDF (only dump created chn)	ChnId	Channel id
  	[c_param] Common parameter	Enable	Enable current channel or not
  		OdBufCnt	The Buffer count that OD stores the result
  		VDFIntvl	Not used currently
  		RstBufSize	The size of the result returned at one time
  	[s_param] Static parameter	inImgW	The Image width input by OD
  		inImgH	The Image height input by OD
  		inImgStride	The Image stride input by OD
  		nClrType	OD input image type 2-y data
  		alpha	Control the learning rate for generating reference images
  		div	Type of OD detection window 0: 1 window 1: 4 windows 2: 9 windows
  		M	How many images update the reference image once
  		MotionSensit	Motion sensitivity setting
  		roi_od_num	Number of boundary points of detection area
  		pnt[i].x	The x coordinate of each boundary point of the detection area
  		pnt[i].y	The y coordinate of each boundary point of the detection area
  	[d_param] Dynamic parameter	thd_tamper	Image difference ratio threshold
  		tamper_blk_thd	Threshold value of the number of occluded areas of the image
  		min_duration	Image difference duration threshold
  VG Attr info for VDF (only dump created chn)	ChnId	Channel id
  	[c_param] Common parameter	Enable	Enable current channel or not
  		VgBufCnt	The Buffer count that VG stores the result
  		VDFIntvl	Not used currently
  		RstBufSize	The size of the result returned at one time
  	[s_param] Static parameter	width	The Image width input by VG
  		height	The Image height input by VG
  		stride	The Image stride input by VG
  		obj_size_thd	The percentage size of the object in the area of interest
  		indoor	Camera mounting area 0-outdoor 1-indoor
  		func_state	Detection mode 2: VG_VIRTUAL_GATE, indicating that the mode is a virtual line segment. 3: VG_REGION_INVASION, indicating that the mode area is invaded.
  		line_number	The number of virtual line segments, support 1 to 4 virtual line segments (The detection mode is effective for VG_VIRTUAL_GATE)
  		p[i].x	X coordinate of line segment n (The detection mode is effective for VG_VIRTUAL_GATE)
  		p[i].y	Y coordinate of line segment n (The detection mode is effective for VG_VIRTUAL_GATE)
  		pd[i].x	The first direction point of line segment n (The detection mode is effective for VG_VIRTUAL_GATE)
  		pd[i].y	The second direction point of line segment n (The detection mode is effective for VG_VIRTUAL_GATE)
  		region_dir	Set the direction of the region invasion. There are currently three directions to choose from: entering, leaving, and crossing. 0-REGION_ENTER: Entering the alarm region triggers an alarm. 1-REGION_LEAVING: Leaving the alarm region triggers an alarm. 2-REGION_CROSS: Crossing the alarm region triggers an alarm. (The detection mode is effective for VG_REGION_INVASION)
  		Pnt[i].x	The x coordinate of the i-th point of the region. (The detection mode is effective for VG_REGION_INVASION)
  		Pnt[i].x	The y coordinate of the i-th point of the region. (The detection mode is effective for VG_REGION_INVASION)
  dump buf status	Channel[xxx]	channel id
  	Input/Output	sequenceNum	The unique identifier of the buffer, the serial number id
  		status	The status of the buffer. (ready: vdf successfully applied for buf; geted: the algorithm successfully obtained buf; drop: abnormal situation, discard processing; done: algorithm operation is successful; del: memory to be released)
  		bufinfo	memory address of buffer
  		getTime	The time it takes for vdf to successfully obtain “task buffer” (us)
  		processTime	The time it takes for vdf to go from enqueueTask to dequeueTask(us)

5.2. echo

# echo help > /proc/mi_modules/mi_vdf/mi_vdf0

Echo help view available commands.

Table 2-1:

Function	Enable/Disable en_workmode(od/md/vg)
Command	echo en_workmode param1 param2 > /proc/mi_modules/mi_vdf/mi_vdf0
Parameter Description	param1: [md/od/vg] param2: [0-1] 0-disable, 1-enable
Example	echo en_workmode md 0 > /proc/mi_modules/mi_vdf/mi_vdf0

Table 2-2:

Function	Enable/Disable vdf chn
Command	echo en_chn param1 param2 > /proc/mi_modules/mi_vdf/mi_vdf0
Parameter Description	param1: [chnId] chnId 0-63 param2: [0-1] 0-disable, 1-enable
Example	echo en_chn 0 0 > /proc/mi_modules/mi_vdf/mi_vdf0 //cat /proc/mi_modules/mi_vdf/mi_vdf0 again, you can see that ch0 of input/output has been closed

Table 2-3:

Function	Dump yuv
Command	echo dump_image [param1, param2, param3] > /proc/mi_modules/mi_vdf/mi_vdf0
Parameter Description	param1: [chnId] chnId 0-15 Param2: [dumpNum] 0-xxxx Param3: [dumpPath] yourPath
Example	echo dump_image 0 10 /mnt > /proc/mi_modules/mi_vdf/mi_vdf0

For example, dump chn0 10 YUV in /mnt directory: