A YOLOv4-tiny Based Moving Object Detection Method for Embedded Vision System

YOLO nets have been widely used in the field of object detection,but the nets are designed for specific object detection in single image without motion information.They are difficult to be applied in embedded vision system directly to do the task of moving object detection.In this paper, combining the YOLO net and inter-frame motion detect,a fast moving object detection method is proposed for embedded vision system.First of all,YOLOv4-tiny is applied to detect objects in sequence images; then,by the the confirmation of inter-frame motion information,every detected object is judged to be moving object or not,the moving objects are saved as final result. Experiments show that the algorithm can detect moving objects in real time and accurately.