Vision Apps User Guide
Stereo-based Obstacle / Freespace Detection Application

Introduction

This application demonstrates obstacle / freespace detection solely uing a disparity map from J7 SDE (Stereo Depth Engine) HWA. It creates disaprity map from a pair of rectified images and determines ground plane. After ground plane estimation, this application detects either obstacles and freespaces with depending on a configuration. In case of obstacle detection, the distances to obstacles are also computed. In case of freespace detection, the maximum drivable distance in front of vehicle is computed as well.

The application makes use of the below application libraries,

  1. Applib for Obstacle / Freesapce Detection using Streo Engine

Supported plaforms

Platform Linux x86_64 Linux+RTOS mode SoC
Support YES YES J721e

Data flow

vx_app_sde_obstacle_detection_data_flow.png

Steps to run the application

  1. Build the application and related libraries as mentioned in Build Instructions
  2. Export path to stereo dataset shown below
    export APP_STEREO_DATA_PATH=/ti/j7presi/stereo_data
  3. Edit "vision_apps/apps/ptk_demos/app_sde_obstacle_detection/config/*.cfg" fields, follow the comments in the app.cfg to understand each field
  4. In the command below, use 'app_obstacle.cfg' for running the obstacle detection functionality and 'app_freespace.cfg' for freespace detection.
  5. Run the app as shown below
    .//vx_app_sde_obstacle_detection --cfg /ti/j7presi/psdk_rtos_auto_j7presi_xx_xx_xx_xx/vision_apps/apps/ptk_demos/app_sde_obstacle_detection/config/<config>.cfg

Sample Output

Shown below are example output

vx_app_sde_obstacle_detection_out.png
Obstacle Detection Output
vx_app_sde_freespace_detection_out.png
Freespace Detection Output