Currently,vision-based road pedestrian detection has become a research hotspot,and for the overlapping field of view demand of joint detection of road pedestrians by vehicle-mounted and roadside cameras,an overlapping field of view pedestrian image alignment method based on the combination of vehicle-mounted and roadside cameras is proposed.Firstly,the YOLOv5 algorithm is used in the roadside unit to measure the specific position of road vehicles in real time,to determine the positional relationship between the vehicle-mounted and roadside cameras,and to improve the accuracy of the subsequent image transformation and parameter optimization;after that,the algorithm combining the mutual informa-tion and gradient is used to align and stitch the pedestrian images in the overlapped field of view.The experimental results show that the average accuracy of vehicle recognition is 88％,and the average response speed of distance measurement is at 46.9 ms.By comparing the alignment errors of different processing algorithms,the image alignment accuracy based on the combination of gradient and mutual information is better than the traditional method.
NETL
NSTL
万方数据
