Ground Obstacle Detection Technology Based on Fusion of RGB-D and Inertial Sensors

Aiming at the real-time detection of ground obstacles in the passable area for the visually impaired, a ground obstacle detection technology based on the fusion of RGB-D and inertial sensor is proposed. Firstly, the spatial model of the ground obstacle is established, and the inertial sensor parameters are fused to calculate the camera inclination to correct the world coordinates of the ground obstacle. Secondly, according to the actual scenes and needs of the visually impaired, the remote detection pixels in the depth image are removed by threshold segmentation algorithm, and the depth map is divided into 4 regions, and the dynamic division of ROI is realized by fusing inertial sensor data. Finally, an obstacle detection algorithm is designed by improved RANSAC algorithm based on the growth of ground area algorithm, and collected real data for experimental verification. The experimental results show that the accuracy and recall rates of the proposed technology reach 90.87% and 89.33%, and better than the existing detection algorithms in the time efficiency of execution, which meets the real-time requirements of the algorithm for the visually impaired.