Research On Perception Technology Based On Fusion Of Millimeter Wave Radar And Vision

Intelligent vehicle is the development trend of automobile in the new era.It is of great significance to improve driving experience,solve traffic jams and reduce traffic accidents.Among them,environmental perception technology as the core part of intelligent vehicle technology is the current research hotspot.At present,single sensor sensing technology cannot fully obtain environmental information.Based on millimeter wave radar and Mobileye,this paper studies multi-sensor information fusion technology to achieve stable and comprehensive environmental information perception,and provide guarantee for subsequent decision-making and planning.Firstly,this paper analyzes the current mainstream sensor sensing fusion technology at home and abroad,compares the advantages and disadvantages of various sensors,proposes a sensing technology framework based on millimeter wave radar and Mobileye fusion,and clarifies the main research contents of this paper.Secondly,according to the requirements of the project,the sensor selection is carried out,and the related performance and parameters are introduced.According to CAN parsing protocol,compile parsing code to receive and analyze sensor CAN data.Analyze the characteristics of millimeter wave radar and Mobileye output signals,set different filtering principles,and preprocess the data.Among them,static and non-collision targets are filtered by setting distance and speed thresholds;the effective life cycle method is used to judge the survival of the target in multiple frames;the Kalman filter is designed to reduce the noise interference.Through the above filtering method,the invalid noise is filtered and the real-time performance of the algorithm is improved.Then,build the fusion algorithm architecture.First,the sensor is synchronized in time and space as the basis of data fusion.The sensor spatial synchronization is realized by using the spatial coordinate system conversion;the linear interpolation method is used to make the high frequency signals downward compatible,so as to ensure the unity of the acquisition frequency of each sensor.Then,the mainstream data association method is studied and analyzed.According to the demand,the Global Nearest Neighbor method is adopted to complete the multi-sensor target matching.The Kalman weighted fusion algorithm is used to fuse the heterogeneous data of the same target,so as to eliminate the data ambiguity.Finally,combined with the underlying self-contained algorithm of the sensor,the tracking strategy is set to achieve multi-target tracking.After that,the driving simulation model is built by using MATLAB automatic driving toolbox.Including the construction of road scene and vehicle target,sensor layout and parameter setting,and the compilation of fusion tracking algorithm.Combined with different road condition information and adding simulation noise,the simulation test is carried out to verify the feasibility of the algorithm.Finally,in order to test the effect of the algorithm under real road conditions,a real vehicle test is carried out.Install millimeter wave radar and Mobileye on JAC i EV3 pure electric vehicles,and perform sensor calibration.Deploy relevant algorithms on the industrial computer for real-vehicle testing.The results show that the method proposed in this paper can stably identify targets such as vehicles and pedestrians under common weather such as rainy and cloudy days,and can achieve multi-target tracking,and has good environmental adaptability;the problem of missed detection of single sensor is solved by using multi-sensor fusion technology,which has good stability.In the stage of environmental awareness technology needs to be improved,the method proposed in this paper has a certain application prospect and research value.