Abstract & Overview
The extensive deployment of video surveillance has underscored the importance of efficient video information processing. Tracking moving objects in video footage constitutes a crucial aspect of this process. The Kernel Correlation Filter (KCF) algorithm is widely employed in the domain of video processing due to its rapid tracking speed and high efficiency. However, its performance can be significantly hindered by environmental factors such as lighting variations, changes in scale and shape, and motion blur. This paper presents an enhanced version of the KCF algorithm, incorporating reliability estimation and a re-detection mechanism to address issues related to severe occlusion and rapid target movement. Additionally, the scale pool method is introduced to improve adaptability. Simulation experiments and real-world video testing demonstrate that the improved KCF algorithm can effectively track moving objects under adverse environmental conditions, yielding satisfactory results.