Customer Detection and Tracking By Deep Learning and Kalman Filter Algorithms

Multiple people's tracking and detecting are essential and challenging tasks for monitoring vision, particularly in public places like shopping centers. We introduced a technical proposal based on the detection algorithm and tracking algorithm. In this investigation, we applied YOLOv3 as a detection algorithm and to raise the performance and velocity Kalman Filter was used as a tracking algorithm. We additionally validated the performance of the proposed detection method “YOLOv3” by performed a statistical analysis of detecting errors MSE method. The experimental results on a series of tests show that the efficiency of the proposed method is quite beneficial, and is able to detect people not only in closed space but in open space, especially when they are closer to each other as well.