Research On Vehicle Critical Information Detection Algorithm System Based On Deep Learning

With the rapid development of China’s economy,the domestic car ownership has increased rapidly,which has brought a series of traffic problems.In order to solve these problems,all the cities in China are developing the intelligent transportation system.The detection of vehicle information is a very important part of the intelligent transportation system,which is of great significance for the development of the domestic urban intelligent transportation system.The effect of the traditional vehicle detection algorithm often can’t achieve the ideal effect.For example,the vehicle detection algorithm based on feature information is not satisfactory for the detection of situations that require complex environment and high real-time requirements.Vehicle detection algorithms based on operational information are not detectable for vehicles that are stationary in the video.At present,the application of deep learning technology in the field of target detection effectively improves the existing problems of traditional algorithms.In this paper,the SSD(Single Shot MultiBox Detector)target detection algorithm is applied to the monitoring video target detection field,and the following work has been completed:(1)Build a deep learning dataset about vehicle information detection using the pictures from a public security bureau.We have selected about 10,000 representative images from the approximately 100,000 images provided as our dataset.Since the data set used is not public,so to do their own sample tags,this article uses github’s open source tool LableImg sample tags.In this paper,we use 7000 images in the dataset as the training set and 3000 images as the test set.(2)In this work,the SSD target detection algorithm is applied to the video surveillance system.The algorithm predicts the detection target by drawing boxes in different features.Since the feature points of different feature layers have different receptive fields,the higher the number of layers,the larger the receptive field of feature points.The low-level feature map is used to detect small targets.Due to the low-level feature map feature extraction is not sufficient,resulting in the algorithm for the detection of small targets inaccurate.In this paper,SSD target detection algorithm is improved,adding deconvolution module in the network to improve the detection accuracy of small targets.(3)Based on the improved algorithm,this paper designs a vehicle monitoring software system,which will help the application of the algorithm.The system interface provides four modules,including the basic system information module,image detection module,real-time monitoring module,history module.