Research on human following technology of robot based on ECO-HC
Qingsen Su, Yanfang Zhang, Peng Wang, Guangliang Liu, Lin Sen Zhu, Wenjie Li · 2023
Due to the poor robustness of the following robot to stably follow the target and relocate the lost target in the dynamic environment, this paper proposes a mobile robot target following method combining YOLOv5 and ECO-HC tracking algorithm. It mainly uses the fast and accurate target detection ability of YOLOv5 to provide the necessary tracking target information for the ECO-HC tracking algorithm. By extracting the FHOG (Histogram of Oriented Gradient) feature and CN (Color Name) feature of the target to train the correlation filter, it is ensured that the following robot can follow the target for a long time. In order to avoid problems with the target loss caused by the block, using DIoU as a criterion for data matching, achieve target recognition. The target detection and tracking algorithm in ROS platform deployed to the mobile robot, and use the Kinectv2 depth camera to obtain follow real coordinates of target. The target following experiment is completed in the experimental scene, and the experimental results show that the mobile robot has strong robustness and can follow the target in real-time.