Research On Vehicle Detection Algorithm Based On Lightweight Neural Network

With the gradual improvement of China’s material living standards,China’s car ownership is also increasing year by year.Although cars bring many conveniences to people’s life and work,they also bring many problems.Such as traffic accidents,traffic congestion,disorderly parking and waste of parking space resources.Therefore,in order to solve these problems,we need to vigorously promote the development of intelligent parking lot system,auxiliary driving system and intelligent transportation system.Vehicle detection technology based on deep learning is a hot research direction in computer vision.It is also a basic and very important technology to promote the development of the above system.However,there are still many deficiencies in the existing vehicle detection models:(1)There are many model parameters,which are difficult to deploy to the mobile terminal;(2)The model accuracy is low due to missing detection of small targets and occluded targets;(3)The model has poor real-time performance in detecting targets.The above three problems have brought great challenges to the application of vehicle detection technology.Therefore,aiming at these three problems,based on the design of lightweight model,two lightweight vehicle detection algorithms are proposed to improve the detection accuracy and speed of vehicle detection model.The specific work is as follows:(1)Vehicle detection algorithm based on Mobile Vit lightweight.In order to improve the detection performance of the model for occluded targets,the Grid Mask image enhancement method is introduced in the data preprocessing stage;In order to improve the feature extraction ability of the model,the improved Mobile Vit is used as the feature extraction network of the model,which combines the advantages of convolutional neural network and transformer network;In order to improve the recognition ability of the model to small targets,panet network is introduced as a multi-scale prediction network;In order to improve the positioning ability of the model to the target,the positioning loss function adopts CDIOU LOSS.(2)Lightweight vehicle detection algorithm based on YOLOX-tiny.In order to improve the detection speed of the model.Firstly,the tiny version of YOLOX network is used as the basic network to realize lightweight;Secondly,two scale FPN pyramid network is introduced into the prediction layer network,and the method of preset a priori frame is used to train the network;Finally,EIOU LOSS is used as the positioning loss function to improve the positioning ability of the model.