Research On Tiny Object Detection And Recognition Algorithm Based On YOLOv5

Object detection is an important research direction in the field of computer vision.As a widely used target detection algorithm at present,yolov5 shows excellent performance in the general target detection data set.However,in the actual vehicle pedestrian detection data set,there are still problems in the detection of small targets by yolov5 due to the small number of pixels and high similarity of small targets at a long distance.Therefore,this paper improves the yolov5 algorithm to improve its detection effect on small targets.The main work is as follows:(1)In view of the imbalance of samples in the data set,data enhancement methods such as stitcher and scale matching are adopted to enrich the diversity of samples,expand the target training set,and increase the proportion of small targets in the training set.(2)Aiming at the problem that the network mechanism is not friendly to small targets,the algorithm is improved from the network structure by adding a detection head for small targets and optimizing the loss function.(3)A vehicle pedestrian detection system is designed and implemented.The experimental results show that the improved model not only meets the real-time requirements,but also outperforms the improved model in small target detection,and improves the detection effect of vehicles and pedestrians in long distance.