Research And Application Of Vehicle Detection Algorithm Based On Traffic Video Data

With the traffic video surveillance network coverage area is growing,relying on artificial is difficult to do all the monitoring points for real-time monitoring,so through the automatic information collection,computer vision technology and information fusion technology to implement intelligent road traffic management,will become the future development trend of intelligent transportation system(ITS).In this paper,based on the intelligent transportation system,the computer vision technology is used to deteet the moving objects in the ground traffic video data and the video satellite data,and the detection,statistics and analysis of the moving vehicle in the region of interest are realized.The main work of this paper includes the following aspects:(1)Analysis and comparison of different feature points of ground traffic video data and video satellite data,choose the appropriate method of data preprocessing,including image grayscale,image filtering and image morphological processing,then a method based on particle filter is proposed to remove the motion noise in the satellite data.(2)The moving vehicle detection algorithm based on GMM and ViBe is analyzed.The two algorithms are compared and analyzed using the ground traffic video data and video satellite data,it is concluded that the GMM algorithm can adapt to the change of light intensity,but it is difficult to realize real-time monitoring,the ViBe algorithm has a small amount of computation and good anti noise performance,and can be used to detect moving vehicles in real time,but the ViBe algorithm uses an image to initialize the background model,which is easy to introduce the"ghost" phenomenon.In this paper,the GMM-ViBe method is proposed to solve the problem of large amount of computation in GMM algorithm and the "ghost" problem of ViBe algorithm.Firstly,the background model of multi frame image is used instead of the single frame image initialization of ViBe algorithm.Then,based on the background model updating method of ViBe algorithm,an adaptive threshold method is proposed,this method can calculate the threshold according to the complexity of the background,and then determine whether the pixel belongs to the background or foreground.This method can solve the problem of large amount of computation and eliminate the "ghost" phenomenon.(3)The extraction and analysis methods of traffic parameters are studied.Based on the vehicle detection method,the traffic parameter factors such as speed,lane occupancy,traffic density and traffic congestion are extracted,which provide practical reference data for practical application.(4)Finally,the GMM-ViBe method is applied to the vehicle real-time monitoring and monitoring data processing module.The development of this module is a combination of civil and military integration of road transport applications in the Qt Creator 4.0.1 integrated development environment,the use of OpenCV3.1.0 computer vision library and GDAL2.1.1 library to achieve.The module realizes data storage and management functions,data preprocessing function,motion vehicle detection function and traffic parameter extraction and comprehensive analysis function.The test data works well and has important practical application value.