Evaluation of Leader–Follower Position and Distance Accuracy and Obstacle Location Using Overhead Camera-Based Planar Homography

Abstract

The leader-follower robot configuration is assessed in this study with an overhead camera vision system. Pixel-based coordinates and perspective distortion impact position and distance measurements in centimeters in the absence of geometric mapping. In order to address this issue, planar homography is used to convert pixel coordinates into arena coordinates. This system takes pictures from an overhead camera, uses HSV color segmentation to detect obstacles, and finds ArUco markers on the robot to determine the marker center coordinates. Four ArUco markers positioned at the arena's corners are used to compute the homography matrix. A 200 x 150 cm indoor arena was used for the experiment, and the camera was placed 190 cm above the ground. Sixteen test points were used to assess the robot's position. The mapping error for obstacles ranges from 0.6094 to 0.8644 cm, while the RMSE values for the x and y axes are 0.176 cm and 0.351 cm, respectively, according to the results. 2.93% is the average leader-follower distance error.