Night Vehicle Identification Based On Machine Vision And Millimeter Wave Radar

With the increase of car ownership year by year,the road traffic safety problem is more and more serious.It has become an inevitable trend for vehicle technology to reduce traffic accidents by setting up advanced driving assistance system or driverless vehicle system.Adaptive driving system,front collision warning system,automatic braking system and other advanced driving assistance systems and driverless systems all need to monitor the vehicle status at night.Therefore,this thesis studies the night-time vehicle recognition technology based on machine vision and millimeter-wave radar,which has important theoretical and practical significance.In this thesis,millimeter wave radar sensors and visual sensors are selected to identify vehicle targets at night.Millimeter wave radar has the characteristics of far detection distance and can work at night.The machine vision can directly perceive the color information of the object.The use of sensor fusion information to identify the target can improve the target recognition rate and improve the robustness of the recognition system.In this thesis,a series of researches on nighttime vehicle recognition technology are carried out based on the machine vision and millimeter-wave radar sensors.The main work is as follows:(1)The method of vehicle recognition and tracking based on machine vision is studied.Firstly,the taillights are extracted according to the halo and brightness features of the taillights,then the taillights are paired and taillights are used to realize the vehicle identification.Finally,Then the improved particle filter algorithm is used to track the taillights of the vehicles hasbeen detected,and vehicles are tracked according to the taillights has been tracked.(2)Research on target selection and tracking of vehicle targets based on millimeter wave radar.First,the radar jamming signal is deleted,then the effective vehicle target is initially selected according to the distance and lane information.Finally,the vehicle target is tracked continuously according to the Calman filtering algorithm.(3)The joint recognition method based on machine vision and millimeter wave radar is studied.Firstly,the camera is calibrated to establish the conversion relation between the radar coordinate system and the image coordinate system.Then the target radar and camera data are used for data fusion in space and time.Finally,the radar information of the primary vehicle target is displayed on the image to establish a region of interest.Vehicle recognition algorithm based machine vision is used in this area for target verification.Building the hardware framework of the test system.Based on Microsoft Visual Studio,MFC and OpenCV software tools,developing a test system software.The experimental results show that the joint recognition method based on machine vision and millimeter wave radar can be used to identify the vehicle at night effectively,and it has better real-time and accuracy.