A Method to Estimate Position Relationship between Pedestrian and Crosswalk Based on YOLCAT++

Abstract

Pedestrian is one of the most important participants in city transportation, at the same time, because of the property of pedestrian, they are the most vulnerable groups in the transportation. Therefore, the regulation for transportation participants should not ignore pedestrian. Recently, with the development of technology, traffic management becomes more and more intelligent. Video surveillance auxiliary equipments are used as a kind of assistive equipment on more and more streets. With the development of neural network, it is used more and more to deal with image classification and recognition problems. There are many excellent algorithms such as RCNN, YOLO, etc., which have been applied in real life to improve efficiency. The application of neural network on video surveillance can significantly improve pedestrian and small motor vehicle traffic violation inspection efficiency. This paper describes a method of using YOCALT++ model to identify pedestrians and zebra crossings and estimate the positional relationship between them.