Object movement highlighting technique using a deep-learning based object detector for effective UAV control

This paper propose a method that highlights the objects, which are moving in real-time, using a deep-learning based detector. The method first detects the objects in real-time, using the state-of-art deep learning object detector named "YOLOv3". After that, for effective Unmanned Aerial Vehicles (UAV) control, the objects, which were detected, will be highlighted by the image difference function. In this part, the most important job is to know which object has moved and how much it has moved. Our study, focus on these two parts. Using the Bounding box size sort, the object bounding boxes will be sorted from small size to big size boxes and with the Pixel winning function, the object boxes will take the pixels that belong to them. At last, with each object's pixel value the percentage of the object's movement will be calculated.