Research On Vehicle Recognition Algorithm Based On Attention Model Network

With the rapid increase in the number of traffic monitoring,the monitoring data of vehicles in urban road traffic has exploded.Vehicle re-identification aims to search for the same vehicle in a non-overlapping field of view camera network,and has become an important research topic in intelligent transportation systems in recent years.Researchers use computer vision technology and deep learning methods to greatly improve the performance of vehicle re-recognition.However,in real life,due to the effects of illumination,angle,object occlusion,and different camera resolutions,vehicle images have complex background information,which poses great challenges to vehicle re-recognition tasks.In response to these problems,this paper improves a vehicle recognition algorithm based on the attention model network.The main work completed includes the following parts:Firstly,in view of the impact of the complex background information of vehicle images in video surveillance on the training of deep learning network models,this paper proposes an optimization algorithm for vehicle re-recognition based on the attention model network.The algorithm can concentrate the parameter weights of the network model on the foreground information of the vehicle picture,that is the body position,suppress the influence of the complex background,and enable the network to learn more discriminative features.Secondly,for many algorithms,after using convolutional neural networks to extract vehicle features,only a simple metric distance is used to sort the results,and high accuracy cannot be guaranteed.This paper proposes a vehicle re-recognition algorithm framework that incorporates a re-ranking optimization algorithm,and establishes a fusion ranking similarity method through multiple query test sets to optimize the final recognition results.Experiments can prove that this method can improve the accuracy of vehicle re-identification.Thirdly,through a large number of experiments on VehicleID and VeRi-776 datasets,the algorithm of this paper can effectively solve the impact of the complex background of vehicle images on the network model.By comparing several other commonly used vehicle re-recognition algorithms,the effectiveness and applicability of the vehicle re-recognition algorithm based on the attention model network proposed in this paper are verified.