Multiple Moving Object Detection and Tracking Across Complex Terrain and Skylines

This paper presents a practical method for detecting and tracking multiple moving objects in cluttered environments, utilizing affordable consumer-grade cameras as input to achieve a low-cost hardware implementation. Tracking objects across the skyline is challenging due to abrupt changes in contrast at surrounding pixels. By combining well-established techniques—such as the Kalman Filter and Hungarian algorithm for tracking, Key-point detection, Lucas-Kanade optical flow for object detection, and RANSAC for egocentric movement estimation and camera motion compensation—this approach delivers accurate, robust, and real-time results using a non-machine learning solution.