Research On Vehicle Detection Based On Millimeter Wave Radar And Machine Vision

At present,when using multi-sensor fusion technology to detect obstacles and vehicles in front,there are few researches on the effective target recognition during cornering and the detection of vehicles in front of the vehicle when the illumination is insufficient at night.The recognition accuracy of the smart vehicle to the effective target is low,increasing the risk of driving.Therefore,multi-sensor based smart car curve detection,nighttime detection and searching for better data fusion algorithms are important ways to improve the robustness,real-time performance and detection accuracy of unmanned vehicles.This thesis mainly includes the following contents:Firstly,the target detection analysis and effective target primaries for the millimeter-wave radar and the camera in the straight and the curve,the radar detection is divided into small areas and analyzed separately in the large detection area,and the camera detection is to analyze the area to be driven predict and simplify the process,the vehicles in this lane will be switched and determined according to the principle of minimum distance.Secondly,the smart car turning model is established,the world coordinate system,the sensor coordinate system and the image coordinate system are established.The relationship between the three coordinate systems is determined through the parameter calibration of the camera and the millimeter-wave radar.The world coordinate point collected by the millimeter-wave radar is projected to the image coordinate system.Then we deal with the image data of daytime and nighttime respectively.According to the different features of routine daylight and infrared images,we apply different algorithms to image denoising,edge segmentation and image enhancement to improve the target recognition.Then,the multi-sensor time-space data fusion is carried out based on the D-S evidence theory,due to the frequency of the millimeter-wave radar and the camera to collect information is different,the data fusion must be carried out in time and space synchronously,and the data in the world coordinate system point transformation into the image coordinate system to form the target contour initially,and then identify and extract the target contour.By comparing the target image extracted by the single sensor and the data fusion,the integrity and validity of the whole target extraction after the fusion are verified.Finally,the test platform for vehicle testing and research is established,including the smart car hardware platform of radar and camera,as well as Halcon image processing and Carsim intelligent vehicle simulation software platform.C + + program is programmed to measure the distance between the inflection point of the smart car and the vehicle in front,and obstacle avoidance trajectory analysis and target extraction and analysis are carried out to verify the sensitivity of the smart car to the accuracy of the front vehicle contour detection after the data fusion.