| Intelligent Transportation System(ITS)is inseparable from people’s life,and the development of artificial intelligence makes intelligent video surveillance system have more extensive applications.In the actual traffic scenario,the detection and tracking of vehicle targets is an important core of intelligent surveillance system,which has become a hot spot of research nowadays.However,in practical applications,there is a wide variety of targets and there are often interference factors such as occlusion,and it is difficult for a single sensor to collect rich information.This thesis addresses the problem of low recognition alignment caused by mutual occlusion between vehicles and external environmental disturbances in intelligent transportation systems,and introduces a data fusion method to perform alignment fusion of data dimension and multi-target data association in spatial dimension for video information obtained from cameras and millimeter wave radar information,which is organized as follows.(1)The thesis firstly describes the working performance of millimeter wave radar and introduces the principle of millimeter wave radar speed and distance measurement.According to its working performance and principle,preprocesses the acquired target signals,eliminates invalid data,reduces data dimensional errors and improves fusion alignment,applies adaptive Kalman filter algorithm to process the target moti其中on information and reduces system random noise.Aiming at the problem of vehicle targets being occluded in the multi-target tracking process of millimeter-wave radar,a trajectory tracking algorithm based on cubic spline function is proposed to complete the radar moving target tracking,which can effectively solve the occlusion problem and improve the tracking efficiency.(2)Aiming at the problems of missed detection,low sensitivity,and incomplete target vehicle information in the process of video image target detection,this thesis first introduces the theory of video moving target extraction.Then thesis establishes a mixed Gaussian model for moving scenes,and uses the weighted sum of multiple Gaussian distributions to the distribution of each pixel in the image sequence is described,so that the system has better robustness.A monocular visual speed measurement and ranging model based on the least squares method is adopted,which can calculate the speed and position information of the current moving target.The license plate information is extracted from the video image,and the calculated speed and distance values can accurately reflect the actual physical driving information of the current vehicle target,which can improve the success rate of later matching.(3)To address the problem of low target alignment in complex environments and high signal-to-noise ratio scenarios,a data fusion method is introduced to fuse the target information obtained by millimeter wave radar and camera.The spatio-temporal alignment of the two sensors is first performed to determine the conversion relationship between the radar and pixel coordinate systems,and the calibration on the timeline is performed by Lagrangian interpolation.A modified Hausdorff algorithm of distance matching is used in the data dimension to calculate the degree of similarity between the data collected by the two sensors,to determine whether they are state descriptions of the same target,and to match the data with high similarity.The successfully matched data are fused with the D-S intelligent data fusion algorithm to complete the fusion of the two sensor data and delineate the Region of Interest(ROI)for target vehicle detection.(4)The experimental platform is built to verify the effectiveness of the fusion detection and tracking algorithm in real traffic scenarios,and a human-computer interaction interface is designed and developed based on this system,which can intuitively obtain information on the current speed,location,target license plate,and target vehicle type of the target,etc.The real-vehicle tests show that the vehicle detection algorithm of millimeter wave radar and video fusion has higher detection accuracy,its adaptation to a wider range of scenarios,and has better robustness and performance compared with a single sensor.Compared with single sensor,it has better robustness and real-time performance,which promotes the development of intelligent road monitoring system. |
