Deep Learning Based Efficient Pedestrian Age Group Estimation and Tracking Techniques for Autonomous Mover

For the application of smart autonomous movers, based on the information of body proportion, in this study a simple and effective processing scheme is developed for age group estimation, distance prediction, and object tracking for pedestrians. By the YOLO-based CNN model for head and body objects detections, the recognition accuracies of age group classification for children and adults are up to 90%. By the age group identification algorithm, the average error rate of the head-to-body ratio and height is estimated to be 3% based on the bounding box information given by YOLO. The predicted error rate of pedestrian's distance by the proposed design is about 3%. For real-time embedded applications, the pedestrian's tracking design is implemented on GPU-based platform, and the frame per second (FPS) rate can be up to 20.