Research On Obstacle Detection Technology Based On Laser Radars Data Fusion

The intelligent vehicle is a vehicle with environmental perception, pathplanning, autonomous decision-making, automatic execution. As an important partof intelligent transportation systems (ITS), intelligent vehicle can play a crucial roleon easing the traffic pressure and creating good traffic environment. Laser radar isan active sensor with far measuring distance, high speed, high precision, and lessaffected by environment and so on. Laser radar can provide sufficient informationfor intelligent vehicle environment perception obstacle detection, to meet detectionaccuracy and real-time requirements, which has important theoretical significanceand practical value.On the research platform of BJTU-IV, this paper studies detection techniquesof obstacle around the vehicle and long-distance in front of the vehicle. For theproblem of machine vision detection results influenced by the external environment,this paper uses laser radar to detection obstacle, to improve the adaptability ofvehicles on the environment. Considering single laser radar’s small scanning rangeand high undetected probability, this paper uses a variety of laser radars to productdata fusion, to achieve a comprehensive, accurate detection effect and meet theneeds of intelligent vehicle safety driving. The main research contents, researchmethods are summarized as follows:Firstly, the paper analyzes the current research status of the intelligent vehicleat home and abroad, research situation of obstacle detection technology by usinglaser radar. For existing deficiencies, this paper presents an idea of data fusion bylaser radars in intelligent vehicle.Secondly, for the issue of laser radar data acquisition and data processing, thispaper writes data acquisition and processing software, achieves data acquisition,preprocessing, clustering, object recognition and tracking algorithms and obtainscommonly used parameters through experiments, to reach real-time effects.Thirdly, for a variety of laser radars optimized deployment and efficient fusionissues, this paper establishes principles of optimal deployment, rational design offusion structure and designs the structure of various laser radars networking anddata fusion, to achieve the full effect of detecting obstacles. After discussing somecommon data fusion algorithm, such as Bayesian inference, Dempster-Shaferalgorithm, voting fusion algorithm, for the detection of obstacles around the vehicle,this paper proposed grid map fusion algorithm based on meshing, mesh clusteringmethod, Mesh barriers occupy theory and grid voting fusion algorithm. For obstacledetection long-distance in front of the vehicle, this paper proposed data fusionalgorithm based on characterization, to achieve multiple levels and accurate detection effects.Finally, this paper build hardware and software test platform of various laserradars data fusion system and conducted a large number of experimentalverification. The results show that the data acquisition software and a variety oflaser radars’ data fusion algorithm have better feasibility, timeliness and reliability.