YOLO-Based Object Detection and Deep SORT Tracking for Remote Sensing Video Analysis

Abstract

Remote sensing video analysis has become an important research area in computer vision due to its wide applications in surveillance, traffic monitoring, and environmental monitoring. Traditional object tracking techniques struggle with complex scenes, occlusion, and dynamic backgrounds present in aerial videos. Recent advances in deep learning have enabled more accurate and efficient object detection and tracking methods.

This research presents a framework for remote sensing video tracking using deep learning techniques. The proposed system integrates the YOLO (You Only Look Once) object detection model with the Deep SORT multi-object tracking algorithm. YOLO is used to detect objects in each frame of the video, while Deep SORT is used to track detected objects across consecutive frames while maintaining consistent identities.

The system processes remote sensing videos by extracting frames, detecting objects, and associating them across frames using motion prediction and appearance feature matching. Experimental results demonstrate that the proposed method improves object detection accuracy and tracking performance in complex environments.

This framework can be applied in several domains including UAV surveillance, traffic monitoring, disaster management, and security systems