Research On Indoor 3D Map Building Method Based On Binocular Stereo Vision For Mobile Robot

As mobile robots in the industrial manufacturing,exploring the unknown environment,human services,military and other fields are widely used,to study how to realize the autonomous mobile robot positioning and navigation has always been a hot issue in the field of robot.The SLAM(Simultaneous Localization and Mapping)technology was proposed by many researchers in 1988 as the key to the autonomy of mobile robots.In SLAM technology,the construction of maps is an important part of it,and its accuracy and precision are the basis for the realization of SLAM technology.At present,in the study of the mobile robot map construction theory and method,the three-dimensional map is more practical than the two-dimensional map,the application is more extensive,the research result are fewer,to develop environment adaptable,efficient and practical method of establishing three-dimensional map has an urgent need.The purpose of this paper is to study the related problems in the construction of three-dimensional maps of mobile robots,and to verify the feasibility of the related algorithms in the indoor environment.For this purpose,the following research is mainly carried out.A 3D map building platform of mobile robot based on binocular vision is designed.This topic deeply analyzes the characteristics of the mobile robot's operating environment,adopt the system integration method to design and debug the mobile robot 3D map construction platform in the indoor environment.Based on the understanding of the motion characteristics and mechanical structure of the platform,the mobile robot platform motion model is established,and using the dead reckoning method to achieve robot relative pose estimation.A 3D point cloud data acquisition method based on stereo vision is studied.In this paper,based on the analysis of camera geometric optical model and distortion model,the algorithm of stereo matching in binocular vision is studied.Emphatically probes into the factors that influence the effect of stereo matching,combined with the actual needs of three-dimensional map construction and obtains the key edge feature in the environment by the method of edge detection,and completes the single frame 3D key feature point cloud data acquisition through binocular distance measurement.The method of point cloud preprocessing and 3D point cloud registration is studied.In this paper,a method of ground reflection feature removal based on least squares method is proposed to solve the problem of rough surface features and smooth ground reflection characteristics in indoor environment in indoor environment.Combined with point cloud filter and statistical analysis method,the pretreatment of 3D point cloud registration is completed.According to the characteristics of point cloud distribution and the accuracy of pose estimation,the normal distribution algorithm is used to complete the mosaic and registration of continuous multi frame point cloud,and a 3D point cloud registration method based on edge features is proposed.In this paper,we compare the actual results in cloud registration between the edge feature point cloud and ordinary sparse point cloud.The validity of the proposed algorithm is verified.In order to verify the feasibility of the 3D map construction method studied in this subject,this topic chooses the typical indoor corridor environment,to control the mobile robot platform in the system,completes the creation of the global three-dimensional map in the indoor environment,and creates the effect analysis and analysis of key influencing factors of the experimental results.The experimental results show that the three-dimensional map of mobile robot constructed in this paper can meet the requirements of autonomous positioning and navigation of mobile robot.