Algorithm Improvement of Pedestrians’ Red-Light Running Snapshot System Based on Image Recognition

Abstract

Pedestrians’ red-light snapping system is applied more and more in traffic control. The snapping technology based on face image recognition is widely used, but there are some defects such as low successful capture rate and small number of simultaneous tracking. In order to improve the capture algorithm based on face recognition technology, the MHT tracking algorithm based on upper body recognition is introduced to improve pedestrian tracking accuracy and system load performance, so that the tracking success rate is increased to 85%, and the number of simultaneous tracking reaches 25 people. At the same time, the image quality is judged by image Laplacian variance, face size and angle, and the face capture algorithm is optimized to improve the snap quality of the close-up photos of the face and improve the system availability.