Research On Environment Reconstruction Of Mobile Robot Based On Binocular Stereo Vision

The level of intelligence of mobile robots is not only the embodiment of national scientific research capabilities,but also the embodiment of the country's comprehensive national strength.The mobile robot can effectively perceive the external environment through the binocular vision system.The three-dimensional reconstruction and path planning based on the perceptual information is the premise for the mobile robot to achieve autonomous travel.The main research content of this thesis is to study the binocular stereo vision system on the FESTO robot experiment platform,and obtain the mobile robot motion environment information through the binocular vision system for 3D reconstruction,which provides the basic environment information for the later path planning.The key part of the three-dimensional reconstruction of binocular stereo vision system is stereo matching.The accuracy of stereo matching directly affects the final result of 3D reconstruction.The existing stereo matching algorithms can be divided into local stereo matching algorithm and global stereo matching algorithm according to the optimized theory method.Compared with the local stereo matching algorithm,the global stereo matching algorithm has higher matching precision,so this thesis adopts global stereo matching algorithm.Although the global stereo matching algorithm has high accuracy in global matching algorithm,it has a general effect on local weak texture region matching in the image.Considering the different environments in which the mobile robot is in the project,the sensitivity of the algorithm to illumination changes must be considered.The algorithm must consider the protection of the image's weak texture area information and the light sensitivity.In this thesis,the internal and external parameters of monocular camera and binocular system are respectively calibrated by zhang zhengyou calibration method to provide data for stereo matching and 3D reconstruction in later stage.When studying the stereo matching algorithms,it is found that the information matching of weak texture region is easy to be lost and the matching result is poor.Superpixels can divide the image into different regions,and strengthen the matching constraints to protect the weak texture region's information.In this thesis,the performance analysis of various existing superpixel segmentation algorithms shows that the Simple Linear Iterative Clustering(SLIC)algorithm is more efficient and flexible.After a thorough study of SLIC algorithm,it is found that it is superpixel segmentation from the initial point and finally reaches the edge of the image.In this thesis,the image's edge is first strengthened and then the SLIC algorithm is used for superpixel segmentation.Therefore,we propose a superpixel segmentation algorithm based on guided filter and carry out experimental verification on the data set.Considering the different illumination and computational efficiency in the moving environment of mobile robots.this thesis combines SLIC algorithm and Census transformation with low photosensitivity to carry out stereo matching.The accuracy of the Census transform matching,the sensitivity to illumination and the accuracy of the matching result with SLIC algorithm are illustrated by experiments.The validity of the method is verified on the data set.Finally,the SLIC and Census transform stereo matching algorithm is used to perform stereo matching on the dataset image,and the dataset is given to the camera parameters to obtain a three-dimensional map.Then,the FESTO robot collects information from the surrounding environment and preprocesses the information,and combines the internal parameters obtained by the binocular camera through the calibrated experiment with SLIC and Census transformation stereo matching algorithm to reconstruct the actual mobile robot environment.