Font Size: a A A

Research On 3D Map Building Method Of Mine Environment Based On Machine Vision

Posted on:2019-08-10Degree:MasterType:Thesis
Country:ChinaCandidate:S Y SunFull Text:PDF
GTID:2428330566963295Subject:Control Science and Engineering
Abstract/Summary:PDF Full Text Request
At present,coal resources play a leading role in China's energy structure.Coal mining process is prone to cause all kinds of mine accidents,resulting in staff trapped or dead.Mine rescue robot can replace rescue personnel for rescue and information detection work,which is of great significance for mine rescue.Robot self localization and 3D map of surrounding environment are the prerequisites for autonomous navigation and rescue detection of mine rescue robots.Robot SLAM mainly realizes robot localization and map building in unknown environment.In this paper,SLAM is applied to the real time map building of rescue environment,and a method of building 3D map of mine environment based on machine vision is proposed.First of all,machine vision information acquisition and preprocessing analysis.Aiming at the influence of noise on the deep image in the mine,a large number of experiments were carried out on the basis of Gauss filtering algorithm,median filtering algorithm and bilateral filtering algorithm.The experimental results show that Gauss filtering algorithm shows good performance in filtering depth image noise,and improves the quality of depth image information at the same time to meet the real-time requirements of 3D map building under the mine environment.Secondly,in order to solve the problem of the failure of image feature matching in the Kinect acquisition of image texture information in the mine environment,a image quality detection–based RN-GM algorithm is proposed based on the analysis of RANSAC algorithm and GMS algorithm.Experiments demonstrate that the RN-GM algorithm has good performance,which can significantly improves the robustness of incremental 3D map building.Finally,in view of the influence of the incremental 3D map,the L-M method is used to optimize the local map and the Bo VW method to achieve the global consistency map,so that the accuracy and consistency of the 3D map are improved.Aiming at the problem that the storage space of 3D point cloud map is large and can not be directly used for navigation,a 3D octree map building method based on machine vision is proposed.The study of 3D map construction in mine environment has important theoretical value and wide application prospect for mine rescue,which will lay the foundation for autonomous detection and rescue of mine rescue robot.
Keywords/Search Tags:machine vision, feature matching, point cloud map, octree map
PDF Full Text Request
Related items