Research On Localization And Mapping Of Mobile Robot Based On Machine Vision

With the comprehensive development of the Industry 4.0,the robot has not only represent the mechanical field,the development of artificial intelligence,machine vision and cloud computing,the robot's function and application scenarios are more abundant,the human society demand for intelligent robot has become diversified,many of those are in demand,the robot can in the space of independent the positioning and three-dimensional reconstruction of this one demand for space is the basis of many functions,but also directly determines whether the robot can appear in the future of human life.The research direction of this paper is that the robot based on RGB-D camera can autonomously locate in space and complete the reconstruction of the scene map.In order to reduce the impact of noise on the system,to make the output more accurate and to improve the stability of the system,this paper mainly studies from the following aspects:Firstly,the captured images of the RGB-D were analyzed,the distribution situation in the depth of phase noise,and then according to the distribution of noise,in order to reduce the influence of noise on motion estimation,a new method was proposed to extract feature points of plane parameters based on noise reduction,the method can will feature the depth value corrected by plane the parameters,thus reduce the influence of noise,finally get a relatively stable motion estimation,the system robustness is improved,compared with the traditional extraction method is also given.Secondly,the SLAM(Simultaneous Localization And Mapping)non parametric equations are discussed,according to the observation data derived form the probability model for maximum likelihood estimation in the least squares model derived equations,the original problem is transformed into a nonlinear problem.At the same time,in the process of solving nonlinear equation,this paper discusses the sparsity of the increment equation,and the sparsity can guarantee the efficiency of the equation and satisfies the real-time requirement of the system.Thirdly,this paper joined the loop detection module in the system,loop detection module is able to reduce the cumulative error of the localization algorithms,only consider data association in continuous time caused by the accumulated error,and loop detection in space constraints considered in,through matching on the shot image the same scene in different time,provide new constraints to achieve the aim of eliminating accumulated error.Finally,the 3D reconstruction of the scene is carried out according to the trajectory data obtained from the final optimization.In this paper,two mapping methods are shown: point cloud map and octree map.According to the different usage scenarios and application requirements,the methods of constructing these two maps are analyzed and discussed.