Research On AGV Autonomous Navigation Based On RGB-D Camera

With the rapid development of industrial automation,more and more intelligent factories were born.Automated Guided Vehicle(AGV)has been widely used in intelligent factories because of its high intelligence,environmental adaptability,reliability and safety.This paper studies the autonomous navigation technology of AGV with Kinect V2 camera.The main research contents are as follows:(1)The imaging model of the Kinect V2 camera was introduced,the causes of radial and tangential distortion of the camera are studied,and the distortion is corrected by polynomial function.The calibration principle of Zhengyou Zhang calibration method is studied,and the internal parameters of color camera and depth camera,and the relative pose between the two cameras are calibrated by this method.Through calibration experiment,it is proved that the misalignment between color image and depth image is improved obviously after the camera calibration.(2)The working principle of the original feature extraction algorithm is studied.Aiming at the problem that the feature points extracted by it are not evenly distributed in the.image,an ORB feature extraction algorithm combining region partition and adaptive threshold value is proposed.To solve the problem that the sparse map based on feature points built by ORB-SLAM2 cannot be used for navigation,the function of stitching global dense point cloud map is added into the map building module of ORB-SLAM2.The point cloud map is transformed into an octree map,and the octree map is projected along the Z axis to get a 2D raster map for navigation.The experimental results show that the ORB feature extraction algorithm combined with region partition and adaptive threshold is more uniform in the distribution of feature points.Compared with the 2D raster map constructed by lidar,the 2D raster map constructed by the improved ORB-SLAM2 for navigation contains more abundant information about obstacles.(3)The principle and problems of A*algorithm are studied.Then an improved A*algorithm is proposed to solve the problems of traditional A*algorithm in path planning,such as long path length,many inflection points and large turning angle.Based on the traditional A*algorithm,this algorithm first considers the turning cost and integrates the pre-judgment planning strategy to plan an initial path with shorter path length,fewer inflection points and smaller turning angle.Then,the redundant inflection point elimination strategy is adopted to further optimize the path length,number of inflection points and turning angle in the initial path,so as to obtain a better driving path for unmanned vehicles.In order to verify the effectiveness of the improved algorithm,a simulation experiment was conducted.The experimental results show that the path planned by the improved A*algorithm has the advantages of short length,less number of inflection points and small turning angle compared with the traditional A*algorithm.(4)The experimental platform of McNum wheeled AGV autonomous navigation equipped with Kinect V2 camera was built,and the visual navigation framework proposed in this paper was given.The navigation trajectory experiment results show that the search time of the improved A*algorithm is slightly increased,while the path length and the number of inflection points are reduced,the turning angle is also smaller,and the smoothness of the path is significantly improved.Obviously,such a path is more suitable for AGV autonomous navigation.The experimental results show that the navigation accuracy of AGV meets the application requirements and has good practical application value.