Flood Rescue Using Multi-object Motion Tracking

Computer vision plays a major role ranging from identifying people to identifying objects in images and videos. Identification can be processed with the object-processing strategies. Image tracking over critical areas aids a lot in this respect. In this project, the concern is detecting the appearance and tracking the motion of persons in a panoramic view and processing the speed of the input images or videos. To create a panoramic view you can stitch videos as input from cameras which creates a larger field of vision. In a panoramic view, the motion of persons can be detected and tracked. Then the count of the detected persons will be updated. The output of our project is a software can be mainly used during the flood rescue which helps the rescue team to get the count of the persons in the flood affected areas. The object detection and recognition can be performed by using Open cv and yolo which is a detector, applying a single neural network which is mainly used for multi label classification. Then the objects are tracked by applying Deep SORT(Simple Online and Real Time Tracking) algorithm for motion and appearance features of the objects with high accuracy.