Research On Scene Mapping Technology Around Robot Based On Stereo Vision

With the increasing level of robot intelligence,the research on autonomous robot navigation has naturally become a hot topic.The key factor for the robot to realize autonomous navigation is the accurate perception and mapping of the environment on the walking path.Only by knowing the surrounding environment can we have a reasonable plan for the robot's movement trajectory,which is also a promotion for unmanned driving technology.At present,the mainstream scene mapping technology has two methods:(1)Multi-sensor combination method for mapping,measuring the physical distance of surrounding obstacles through multiple sensors,and using a large amount of distance information to restore the scene.This method is costly and difficult to restore.Scene appearance.(2)Use camera shooting for scene mapping.This method intuitively presents the two-dimensional picture information of the scene,and is assisted by algorithms to make the scene mapped to the three-dimensional space have high environmental integrity,which provides a stable guarantee for the accuracy of the robot's walking path.In this paper,the method of binocular stereo vision is used to map the surrounding scene of the mobile robot.The principle of binocular stereo vision is essentially a simulation of the principle of human vision.On the basis of the parallax theory,three-dimensional restoration of the scene collected by the camera is performed.Therefore,the most critical part of binocular stereo vision is to obtain the disparity map.The quality of the disparity map directly determines the effect of the 3D reconstruction,which is also the focus of this article.In order to obtain a high-quality disparity map,this paper will experiment and analyze the three typical algorithms of local stereo matching algorithm,semi-global stereo matching algorithm,and global stereo matching.The method of using control variables will adjust the parameters that affect each algorithm,and select a disparity map with the best effect.Based on comprehensive considerations,the SGBM semi-global stereo matching algorithm was finally selected as the algorithm for obtaining the disparity map in this subject.Aiming at the black hole area in the disparity map,an adaptive window average filtering algorithm is proposed to fill the disparity black hole,making the disparity map more complete.In this paper,the internal parameters,external parameters and distortion coefficient of the camera are obtained through camera calibration.The binocular camera collects pictures of the scene where the robot is walking along the way and performs stereo correction.In the experiment,due to factors such as camera pixels and uneven illumination,the quality of the collected photos is not high.In order to solve the problem,the method of histogram equalization is used in the image preprocessing stage to enhance the image to be matched.A stereo matching algorithm is used to obtain a disparity map from the enhanced image group pair,and the disparity map black hole region is optimized and filled.The combination of disparity map,camera calibration parameters,texture mapping,and triangulation transformation principle can restore the scene collected by the camera into a three-dimensional point cloud in space.Control the robot to walk,continuously use the binocular camera to collect the surrounding scene state along the way,and restore the 3D point cloud information of the scene in the current state every time it collects.Through the relevance of the point cloud feature information in adjacent states,point cloud splicing is carried out in turn,In this way,the surrounding information on the walking path of the robot is reconstructed,and the task of mapping the surrounding scene is completed.