Research On The Vehicle Detection And Tracking Algorithm Based On Deep Learning

With the rapid popularization of 5G and the rapid development of the artificial intelligence industry,unmanned driving technology has become a new and hot research direction,and vehicle detection and tracking are one of the important research hotspots in the current unmanned driving field.Due to the complexity of the real road scene and the existence of interference factors such as occlusion,the research of vehicle detection and tracking algorithms is facing severe challenges.How to achieve rapid and accurate detection and tracking of vehicles is an urgent technical problem to be solved.This paper studies the vehicle detection and tracking algorithm based on deep learning.The main work content is as follows:In terms of vehicle detection,the current widely used target detection algorithm Faster R-CNN is first studied.Aiming at the slow detection speed of Faster R-CNN algorithm and low detection accuracy of small targets,an improvement plan is proposed.By migrating the lightweight network model shuffle-net to the Faster R-CNN target detection framework,the amount of model parameters and calculations are greatly reduced,thereby improving the detection speed.At the same time,by adding a deconvolution operation after the multi-scale feature fusion,the rich information in the shallow layer is fully utilized to strengthen the feature performance of the decision layer.Through the fusion of the output feature information of different convolutional layers,the complete reflection of the small target information is realized,thereby improving the overall detection accuracy of the algorithm.In terms of vehicle tracking,this paper makes full use of the advantages of neural networks and proposes a tracking algorithm based on block anti-occlusion,which realizes the stable tracking of vehicles.The research method organically combines the convolutional neural network model and the relevant filtering framework,and uses a smaller-scale convolutional neural network to extract the features of the tracked target signal in blocks.By performing occlusion and scoring processing on the part of the block,and predicting and estimating the position of the tracking target under the framework of the relevant filtering model,the tracking performance of the algorithm when the target is occluded is effectively improved.The tracking performance of the proposed anti-occlusion tracking algorithm based on block is tested on the OTB data set.Compared with other tracking algorithms,the tracking algorithm proposed in this paper can effectively suppress the influence of interference factors and achieve high-precision and robust vehicle tracking.