Research On Key Technologies Of Millimeter Wave Radar And Machine Vision Dual Mode Detection

With the development of science,technology and socitey,intelligent driving has become the development trend of traffic in the future.At the same time of improving the active safety performance of the car,improving the efficiency of traffic,and improving the driving experience.Perception is the premise of intelligent driving,and the eye of intelligent car.The sensors used for perception mainly include camera,millimeter wave radar,ultrasonic radar and lidar,among which MMW radar and camera are two key technologies.MMW radar has many advantages in target detection,such as long detection distance,accurate measurement results,strong anti-jamming ability,etc.,but it is a major defect of radar that target recognition information cannot be obtained.Vision sensor does not have this defect,and with the development of deep learning,the accuracy of identifying vehicles,pedestrians and other targets is greatly improved,but it is vulnerable to the influence of light conditions,and the applicability is not very stable.The traffic environment is becoming complex,and the use of multi-sensor technology is an important technical means for the development of intelligent driving and one of the key development directions of vehicle active safety.Based on the research background of intelligent driving,this paper studies and improves the target detection methods based on millimeter wave radar and machine vision respectively,and proposes the vehicle forward sensing detection method based on the fusion of millimeter wave radar and machine vision target.The main research contents are as follows:1.Based on SG24T1 and SG24R1,the front-end hardware circuit of 24 GHz MMW radar is designed.A microstrip antenna array for 24 GHz radar is designed.The key indexes of the front-end and antenna meet the requirements of 24 GHz radar.The radar board has been processed and can detect the target after testing.2.Based on zynq-7020,the tiny-YOLO neural network is transplanted,and the deep learning network model is segmented to hardware acceleration in the embedded system.The target is detected and recognized.The detection speed is real-time and the power consumption is low.Through the camera calibration to achieve the measurement of the target distance,the target position of the pixel coordinate system is mapped to the world coordinate system to achieve the unity of the coordinate systems.3.Based on Hungarian algorithm and Extended Kalman filter algorithm,the target level fusion of MMW radar detection target and machine vision detection target is designed.The fusion algorithm can effectively reduce the false detection of MMW radar and the missed detection of machine vision detection,and the fusion speed is fast.From the perspective of practical application,this paper studies the key technologies of MMW radar and machine vision for target detection,and designs a fusion algorithm,which improves the redundancy and fault tolerance of the system,and has certain significance for the application of intelligent driving and active safety.