Research On Obstacle Detection And Tracking Based On The Single Laser Radar

The research of the unmanned ground vehicle is one of the most important part ofthe artificial intelligent. And environmental perception is a key module in theunmanned vehicles.At this stage, a lot of research on obstacle detection and trackinghave been done at home and abroad, but now the algorithm about it is still imperfect,data processing is more complex, and the equipment is expensive. To solve theseproblems, based on single laser radar ranging technology, combined with the field test,this paper proposes a new improved obstacle detection and tracking algorithm, andrealizes the unmanned vehicle detection and tracking of obstacles.The main work of this paper includes:(1)Based on the surface characteristics of different types of obstacles, and thecharacteristics of the single laser radar, this paper proposes a adaptive thresholdnearest neighbor clustering algorithm.The median filter is used to remove the noiseconsidering the characteristics of the single scan lidar echo data. Then making use ofthe distance of the adjacent two points to cluster obstacles.At last the effectiveness ofthe algorithm is verified by an experiment.(2)Because the clustering criterion is single, combining the intensity of laserreflection mechanism, the feasibility of the reflection strength criterion is validatedwith experiment.Combined with distance and intensity of reflection to improveclustering algorithm and to improve the precision of the obstacle clustering. Then theexperiment verifies the feasibility of the criterion.(3)For the target tracking, the advantages of the kalman filter and particle filteralgorithms are discussed. This paper proposes the nearest match tracking methodusing the kalman filter for reference. By comparing the distance of different targets inadjacent time, the same target’s location in adjacent time is obtained. And obtain therelative velocity and moving direction of obstacles of target information, etc.(4)The algorithm is realized using C++and Qt programming language. Theresults of experiments, both in simple indoor environment and complex outdoorenvironment, prove the feasibility and effectiveness of the algorithm.