This study aims to develop a collision avoidance system that detects objects using a stereo camera for a robotic combine’s safe operation. Herein, two methods for object detection and distance estimation have been poposed. The first method is a combination of a deep learning-based object detection method, i.e., you look only once (YOLO), and a depth image, whereas the second relies on point cloud data, which are obtained from the stereo vision. Results show that both the detection methods achieved an F-measure of more than 0.95, however, YOLO-based method needs a high performance computer. Stereo vision system could be applied to real-time object detection and collision avoidance for a robotic combine, which can take actions, slow-down, and stop, according to the position of the detected obstacle.
View full abstract