Environmental Perception Of Excavating Robot Based On Multisensor Fusion

In order to excavate the robot to perceive the complicated environmental information from the field,this paper mainly studies in detecting obstacles,segmenting and establishing map in unknown environment.In order to segment obstacles in the field quickly,this paper proposes a watershed algorithm based on the lidar information and the regional growth.The laser radar information algorithm is used to guide the segmentation area.Combining the regional growth and watershed algorithm and utilizing the information of the lidar is to determine the optimal threshold of the regional growth.The results of the experiences has showed that the method can effectively overcome the over-segmentation of the watershed algorithm and meet the real-time requirements.When operating,the excavating robot is supposed to segment targets accurately.In this paper,we propose an algorithm,known as Grab Cut image,which can adapt the background without any instruction.Neighboring pixels as background pixels to complete the interaction of the Grabcut algorithm by making background become simple,reducing the number of nodes in the background model,and using the information of the laser radar.The experimental results has showed that the improved algorithm can improve the running efficiency well.The edge of the target information is more complete and smooth.Aiming at the problem that the monocular camera may only acquire a part of the information of obstacles,this paper adopts the binocular image splicing technique.The gradual integration of progressive algorithm always comes with questions like unclearness and overlapping.In order to deal with these problems,we have proposed to use Sobel edge detection and the image edge difference to correct the fusion region.The experimental results has showed that the methodcan make the fused images clearer.Then,this paper introduces a raster map,which uses obstacles of the previous frame image as a landmark to match the local map created by the back frame image to update the global map,the laser radar related to the back frame image and the laser radar to locate the distance and angle of the obstacle in the back frame image,in some detail.Finally,this paper proves the feasibility of SLAM map construction which is based on extended Kalman filter.