Vehicle Detection And Tracking Base On The Intelligent Monitoring

The detection and tracking of Moving vehicle is one of the core parts of modern intelligent transportation system.Based on the analysis and summary of the existing vehicle detection and tracking technology,the paper focuses on the motor vehicle automatic detection,shadow removal,vehicle tracking technology,and proposes the improved algorithm.The main research work as follows:(1)The research of Moving vehicle detection technologyFor common vehicle detection method is more sensitive to light and high requirement of contrast values between vehicle and poor background,the paper put forward a method that we use similarity of LBP texture analysis between background image and the current frame image to extract the prospects:the paper use Gaussian mixture model to obtain the current frame corresponding background frame,calculate the N-LBP texture,and complete the texture similarity analysis between current frame and background frame.Finally,we detect the vehicle accurately and these problems can be solved effectively.(2)Shadow detection of moving vehiclesThe traditional shadow detection methods tend to use a single method for shadow detection,which shadow can be identified mistakenly easily.We propose a new method which combines color with model of light and correlation between neighborhood shadow pixels.Firstly,we use the color characteristics to complete a preliminary inspection of the shadow.Then we use light intensity ratio confidence interval to eliminate the mistakenly identified shadow pixel.Finally,we use the correlation of pixels neighborhood shadow pixels to eliminate mistakenly identified shadow pixels completely.We use the method to ensure the integrity and accuracy of the shaded area detection,so as to achieve the purpose of accurate detection of vehicle.(3)The research of Moving vehicle tracking technologyBecause the traditional Camshift algorithm tracking accuracy and the stability is not high,we put forward a way which use image characteristics,and the combination of the probability of space motion to improve the Camshift algorithm.Firstly,we increase the strength of character by using multi-dimensional histograms of color and texture.To a certain extent,it solves the background colors and target tracking error produced by similarity.Then we introduce some changes in target location prediction mechanism which use L-K sparse optical flow method based on pyramid to estimate the moving vehicle center.It can further reduce the search time Camshift tracking and reduce the amount of calculation.Finally,the traditional reverse projection Camshift algorithm was improved by joining the probability information of space movement pixel.It ensures the stability and the accuracy of tracking in a variety of scenarios.(4)Using the vehicle detection and tracking of results,we successfully obtain the corresponding traffic parameters and analyze the vehicle behavior.