Font Size: a A A

Research And Application Of Vehicle Detection Method Based On Deep Learning

Posted on:2021-02-06Degree:MasterType:Thesis
Country:ChinaCandidate:D XieFull Text:PDF
GTID:2492306305972049Subject:Circuits and Systems
Abstract/Summary:PDF Full Text Request
With the rapid development of national economy and the continuous improvement of people’s living standards,the number of motor vehicles has increased dramatically.The problem of traffic congestion and parking difficulty has become a major social problem to be solved in large and medium-sized cities.Road traffic accidents and road traffic violations occur frequently,and road traffic management,motor vehicle management and driver management are facing severe challenges.Artificial intelligence technology continues to achieve technological breakthroughs,and deep learning has become a hot technology of research and application.It has great academic research significance and engineering application value to carry out application research around vehicle detection method based on in-depth learning.Whether it is the confirmation and punishment of motor vehicle violations,or the gate release and parking charge management of motor vehicles in and out of the park,it is necessary to carry out intelligent monitoring on the behavior state of motor vehicles.The key tasks of intelligent monitoring are the detection and recognition of motor vehicles and the accurate recognition of license plates.This paper focuses on the intelligent monitoring method of motor vehicle behavior state,on the basis of understanding the research status and development trends at home and abroad,learning the basic knowledge of neural network and the common network structure of deep learning,has been carried out a series of research work based on deep learning method.Firstly,a vehicle detection algorithm based on convolution neural network and cyclic neural network is proposed.Using residual connection to improve ResNet-50 network,and using KITTI dataset to train and evaluate the algorithm model.Secondly,the paper proposes the recognition algorithm of indefinite length license plate based on convolution neural network,cyclic neural network and CTC(Connectionist Temporal Classfication)algorithm.Train and test the license plate recognition algorithm by using the dataset generated by the self-designed license plate generator.Thirdly,based on the vehicle detection algorithm,license plate recognition algorithm and multi label classifier,with the intelligent monitoring method of motor vehicle behavior as the main line,seven industry application systems are planned,and the five layer core common architecture of the application system is designed.Fourth,from the perspective of practical application,this paper explores and analyzes several typical application cases of motor vehicle behavior monitoring methods.
Keywords/Search Tags:deep learning, vehicle detection, license plate recognition, multi-label learning, vehicle behavior state
PDF Full Text Request
Related items