A Vehicle Detection Algorithm based on Superpixels Segementation
With the increasing number of landing motor vehicles in China,the road traffic pres-sure has gradually increased.Therefore,it is of great significance to apply artificial intelligence method to quickly identify vehicles and obtain information such as vehicle and road conditions.SLIC uses a small number of superpixels to replace a large number of pixels,expressing image feature,capturing local image similarities,and highlighting the boundaries of each content in the image,thus reducing the complexity of image post-processing to a certain extent.However,the SLIC has a poor effect when the target is extremely similar to the ambient color or bright-ness.In this paper,considering the symmetry and contour characteristics of vehicles,a vehicle detection algorithm based on adaptive superpixels segmentation is proposed to improve the SLIC by introducing variables to measure the importance between brightness and color.This article se-lected BIT-vehicle data set for experiments in YOLOv5.The algorithm can effectively improve the performance of YOLOv5 in vehicle recognition,and the experimental results show that compared with the unprocessed model,the accuracy is improved by 1％,and the mAP is im-porved by 1.3％.
NETL
NSTL
万方数据
