Research And Implementation Of VSLAM Algorithm Based On Graph Matching

With the rise of the intelligent robot industry,the demand for robotics has spawned.For mobile robots,the real time mapping and positioning functions complement each other and are indispensable.SLAM(Simultaneous Location and Mapping)is the key technology to solve these problems.At present,the mainstream SLAM technology has two directions:laser SLAM and visual SLAM.Among them,visual SLAM,which contains four core modules:front end,back end,loop closure detection,and map construction,is widely used because of its low price and high reliability.The focus of this article is on the front end,namely Visual Odometry(VO).In the current SLAM system,the simple matching algorithm of the VO module has large deviations in the key frame matching when illumination change,object occlusion,and motion blur.In this paper,we propose a new improved algorithm by the research of these above problems and conduct experiments on the mainstream framework.The graph matching problem refers to optimizing the consistency relationship be-tween two graph nodes.Graph matching,which is the basis of solving many problems,is widely used in the field of computer vision,such as behavior analysis,target tracking and so on.Based on the ORB(Oriented FAST and Rotated BRIEF)feature points,we use the graph matching algorithm to fully exploit the structure information of the two frames before and after the visual odometry catching.In this paper,the ORB-SLAM2 algorithm framework is developed and improved,n which graph matching algorithm is added to make the system have high accuracy and robustness.The main theoretical basis and research contents of this paper are as follows:(1)A VO algorithm based on graph matching is proposed.The algorithm proposed in this paper combines the excellent anti-interference of graph matching,effectively trans-forms the structure information in the image into graph structure information,integrates the key point structure information and descriptor information and establishes the corre-sponding relationship,and proposes a fusion strategy.(2)The algorithm is applied to ORB-SLAM2 algorithm framework.The ORB-SLAM2 algorithm framework is the mainstream open source framework in the SLAM system,including all the modules of the SLAM.Based on this,the visual odometry part is modified and the graph matching algorithm part is added to improve the current and last frame matching process.Meanwhile,we propose a VSLAM algorithm based on graph matching.(3)Experimental verifications were carried out on multiple public datasets.The VSLAM algorithm based on graph matching was tested on the TUM,KITTI,and EuRoC databases and compared with other algorithms.The experimental results show that the proposed algorithm provides a more accurate and robust trajectory under normal conditions,and has obvious improvement effect,which proves the effectiveness of our algorithm in comprehensive evaluation.