Fast RGB-D ICP With Bionic Vision Depth Perception Model

Mobile robot is an important direction in the field of robotics, how to deal withSimultaneous Localization and Map Building under unknown environment is a keyissue in the field of mobile robot navigation. Based on the previous research andanalysis of SLAM algorithm, we can find that there are different kinds SLAMalgorithm, the choice of SLAM algorithm depends on the hardware conditions ofsensor. As RGB-D sensors appeared in2010, for its high cost performance RGB-Dsensors are widely used in the mobile robot navigation and positioning or other fields.Among them, the new RGB-D SLAM algorithm has always been a hotspot in the fieldof3D SLAM in recent years.In the paper, a bionic vision depth perception model is put forward aimed atreal-time performance of RGB-D SLAM, which takes the sensors’s depth value as aparameter. As we all know biological vision is more clear as the distance closer andmore obscure as the distance further, scene over the depth of field can be negligible.According to that a gradient filter algorithm of point cloud based on sensors’s depthvalue is researched, which can reduce the calculated cost of ICP(Iterative Closest Point)and improve RGB-D SLAM efficiency to get high quality of3D map.Firstly this article summarizes the main content of this topic research the RGB-DSLAM real-time performance, and then sums up the research situation of scholars bothdomestic and international. Then according to the combination of sensor parametersand ICP algorithm RGB-D bionic vision depth perception model is put forward. Inorder to further verify the bionic visual model to feasibility influence of the RGB-DSLAM algorithm, we use the open source ROS (Robot Operating System) as softwareplatform, the omni-directional wheel mobile Robot as the experimental Robot, havinga verification, testing, analysis and evaluation on the improved algorithm performance.Experiments on the mapping accuracy is introduced with Kinect-IMU fusion algorithm to assess the effect of the fusion algorithm separately. Finally in the three-dimensionalglobal figure the bionic visual model and Kinect-IMU two kinds of improvedalgorithm were carried out, the bionic vision model algorithm improved the real-timeof RGB-D SLAM and Kinect-IMU fusion algorithm improved its accuracy.