基于三帧差分法与质心匹配的行人自动计数算法

ALGORITHMS OF PEDESTRAIN DETECTION AND TRACKING BASED ON THREE FRAME DIFFERENCE AND PEDESTRIAN CENTROID MATCHING

为了实现对视频序列中人流量进行实时、准确的检测与跟踪,提出了一种基于三帧差分法的行人质心匹配的行人检测与跟踪算法.采用改进的三帧差分法实现对运动目标信息的检测,然后根据运动目标的信息进行行人面积与质心的计算,并根据质心最近距离匹配、运动方向一致性、越界检测的原则对行人进行跟踪.实验结果表明,该算法在不降低检测效果的情况下明显提高了行人检测与追踪的效率.

To realize the accurate detection and tracking of pedestrian in the video sequence in real-time, a centroid of the pedestrian detection and tracking algorithm based on three frame difference method is presented. Firstly an improved three frame difference method is used to detect moving target information, and then the pedes- trian area and centroid according to moving target information is calculated, and the pedestrian is finally tracked ac- cording to principles of matching of minimum distance, the consistency of the direction of movement, and cross bor- der detection. The experimental results show that this algorithm can significantly improve the efficiency of pedestrian detection and tracking without reducing the detection accuracy, and has a strong robustness.