Multi-target Detection And Tracking Of Mm-wave Radar And Visual Information Fusion

The emergence of intelligent vehicles is an important solution to solve urban road congestion and reduce traffic accidents,and is the main direction of the future development of the automotive industry.In this paper,the method of millimeter-wave radar and visual information fusion is used to output the position information of detecting obstacles around the car,so as to obtain the environmental conditions around the car body and provide a basis for the control and decision-making of the car body.Firstly,the current situation of vehicle sensor fusion at home and abroad is analyzed,and the shortcomings that are easy to occur in the process of the fusion of millimeter wave radar and visual camera are compared and summarized,and the innovation points of this paper are established.According to the innovation points,the solution of this paper is proposed.Then,the sampling principle of millimeter-wave radar is analyzed.The edge operator detection of multiple sets of data photos provides a theoretical basis for sensor fusion,and the selection is established according to the camera performance.Secondly,according to the Haar-like matrix features,vehicle detection is trained.The improved JPDA was selected as the data detection network.The spatial fusion model is established to ensure the spatial identity of the fusion results of millimeter-wave radar and visual camera,and the time fusion of sensors is realized by using the method of interpolation and extrapolation to ensure the consistency of the output results.According to the detection frame intersection ratio and JPDA data correlation algorithm,a fusion strategy is proposed.Finally,the two sensors are calibrated respectively,and a fusion detection algorithm system is built on the real vehicle based on Open CV and VS 2019 C++ platforms.The fusion algorithm is verified,and the recognition accuracy and robustness of the real vehicle data analysis algorithm are obtained.