A Method of Pedestrians Counting Based on Deep Learning
Yile Yang, Weiwei Gao · 2019 3rd International Conference on Electronic Information Technology and Computer Engineering (EITCE) · 2019
In order to realize real-time and accurate population statistics of pedestrians in video, a counting method based on deep learning is proposed. Firstly, the anchor boxes of the detection model is optimized by k-means clustering method. Secondly, the deep learning model YOLO-v3 is trained by pedestrian images obtained in real scenes. Then Deep sort online multi-object tracking algorithm is used to track multiple pedestrians and obtain their trajectories respectively. Finally, the number of pedestrians is determined by the counting line. The counting accuracy of this method is up to 89.2%, and the detection time of each frame is up to 65ms. In addition, it has strong scene adaptability and good robustness, which can meet the requirements of real-time pedestrian counting.