Pedestrian Tracking Efficiency Through Evaluation of Various Optical Flow Algorithms

This study investigates the accuracy of optical flow algorithms-Lucas-Kanade, Farneback, and Horn-Schunck-in motion analysis applications. Significant differences in tracking success rates among the algorithms were analyzed using the chi-square test of independence. Lucas-Kanade demonstrated the highest observed success rate with 87 out of 99 tracked pedestrians, followed by Farneback and Horn-Schunck, each with 84 out of 99 tracked pedestrians. The calculated chi-square statistic of 6.00 exceeded the critical value - 5.991, where the significance level is set at 0.05, which led to the rejection of the null hypothesis. Results indicate a statistically significant association between the algorithms' tracking accuracy and their observed success rates. These findings underscore Lucas-Kanade's advantage in achieving superior tracking accuracy under controlled experimental conditions. The importance of algorithm selection based on the requirements of the specific application is emphasized in the study, highlighting Lucas-Kanade's potential for optimizing motion analysis systems in domains such as robotics, autonomous vehicles, and surveillance. Future research could explore further optimizations to enhance algorithm performance across diverse environmental conditions and motion complexities.