Expressway thrown object detection method based on cascade difference perception model

The invention discloses an expressway thrown object detection method based on a cascade difference perception model, and the method comprises the following steps: carrying out the background modeling of an expressway video, extracting a foreground, and carrying out the mathematical morphology processing, and obtaining a foreground candidate box; removing foreground candidate frames which are too small and too large in area, are not on the road surface and are identified as people and vehicles by the YOLO network model; using an IOU target tracking algorithm to solve a foreground candidate frame which is still for more than 2s; inputting the median image of the current frame and the background image into a difference perception model to obtain a difference region and generate a difference box; retaining a foreground candidate frame matched with the difference frame through an optimization algorithm; and inputting the foreground candidate frame image into a road surface and non-road surface binary classificatio