Research On Regional Navigation Planning Algorithm Of Target Detection Based On Sensor Fusion

As an intelligent automation system combining software and hardware,there are still many problems to be solved in the promotion and application of of self-driving car,however,the research and development of automatic parking technology has more reliable practical application value and technical reference significance.As the key technology of automatic parking,target detection technology based on sensor fusion still has the problems of under realtime detection system and output periodic fluctuation;path planning technology only considers the optimal distance or time,not the difficulty of vehicle handling.This topic comes from the internship project,self guided parking project,the author is mainly responsible for the multisensor data fusion target detection module and regional path planning algorithm design,research and innovation for these problems.Based on the ESR radar and 1080 p wide-angle camera,the target in front of the vehicle is identified,and the sensor detection information is obtained.The multi-sensor data fusion technology is used to improve the real-time and accuracy of the detection system.Based on the parallel data fusion scheme,the joint calibration is completed based on Zhang Zhengyou’s single strain calibration method combined with fish eye model distortion correction.The sampling time deviation is corrected by using Lagrange interpolation and Bayesian estimation.The camera target matchs with radar target image projection and update data based on IOU non maximum suppression algorithm.the tracking management of the detected target is carried out based on the life cycle method.The designed fusion algorithm can reduce the impact of distance on target matching.The fixed output cycle is 25 ms,and the output frame rate is increased to 40 frames,and more accurate real-time target detection results are obtained.The designed path planning algorithm obtains real-time vehicle coordinate information provided by GNSS and constructs high-precision electronic map in Open DRIVE format.Based on the target detection results of sensor fusion,project it into Frenet coordinate system for decision-making judgment,carry out deceleration parking,path keeping,detour or obstacle avoidance planning.According to the decision-making,based on Floyd-A-star algorithm and artificial potential field algorithm,the control difficulty factor is designed to plan the optimal navigation path in real time,so as to ensure that the vehicle can reach the destination quickly and safely.Based on MATLAB / Simulink and real vehicle test,the feasibility and reliability of the algorithm are verified.The results show that the target detection module and path planning module in this paper are designed to obtain real-time and accurate detection information,make reasonable decision to plan the navigation path,and complete the autonomous parking in and out of the point-to-point area in combination with the automatic parking system.As a pre research sub project of the whole project,this research provides a good basic framework for the follow-up function development and optimization,and has great practical application value.