Performance Comparison of KLT and CAMSHIFT Algorithms for Video Object Tracking

Abstract

Human detection and tracking is one of the most crucial tasks in video analysis. We can find its applications in areas like video surveillance, augmented reality, traffic supervision. KLT and CAMSHIFT are two popular algorithms for this task. In this paper, we present a comparison of their performance in different scenarios. As a result, this paper provides concrete statistics to choose an appropriate algorithm for tracking, given the nature of the objects and surrounding. Our experiments show that KLT algorithm is advantageous for crowded scenes, whereas CAMSHIFT performs better for tracking a specific target. Based on our analysis, we conclude that KLT algorithm performs more efficiently than CAMSHIFT algorithm for video object tracking.