| With the rapid development of robotics and artificial intelligence,research in the field of mobile robots is gaining more and more attention.Simultaneous Localization and Mapping is one of the most difficult and attention problems in the field of mobile robots,is the key technology to achieve truly autonomous for robots.In an unknown environment,SLAM technology allows the robot to build a map that is consistent with the surrounding environment by using the data acquired by the sensor,and at the same time to determine the robot's pose.In recent years,due to the unique advantages of the vision sensor, SLAM research based on monocular vision has become one of the important research directions in the field of SLAM.Monocular vision SLAM is divided into a front-end and a back-end,the front-end detects keypoints and extracts descriptors in images,matches the descriptors to previous,according to the matching result,tracks the camera's pose,then identifies the place; According to the results of the front-end algorithm,the back-end optimize the pose graph and build a map.But current monocular vision SLAM is inefficient and can not meet the real-time requirements,and it has a low accuracy,the resulting robot pose and trajectory will generally drift with respect to the real ones.Aiming at solving the problems mentioned above,this paper proposed the following improved methods:(1)Use ORB based method in the feature detection and descriptor extractor stage.(2)Use the optimized feature matching method based on FLANN in the feature matching stage.(3)Use the map initialization algorithm based on contrast and pose tracking algorithm combined with EPnP in the pose tracking stage.This paper uses a benchmark to evaluate the original algorithm and the improved algorithm,of which the result show that improved methods can not only meet the real-time requirement,but also greatly reduce the errors and improve the accuracy.This proves the correcthess of the improvement method proposed in this paper. |
PDF Full Text Request