Research On Visual SLAM For Mobile Robot Based On Kinect

Intelligent mobile robots are required to enable autonomous navigation and positioning in complex environments,simultaneous localization and mapping(SLAM)is a basic and essential technology of the fully autonomous mobile robot.SLAM technology based on vision has been widely concerned by researchers because of its low cost,abundant information and easy to extract features.As Kinect can obtain the RGB-D information of the environment conveniently and quickly,it is widely used in visual SLAM.At present,the mainstream SLAM systems with RGB-D sensor are constructed by front-end of image processing and back-end of pose graph optimization.Aiming at the problem of low real-time performance of the visual SLAM system,we mainly study the key factors about visual SLAM system in the front-end of image processing in this paper.The efficiency of the front-end will directly affect the real-time performance of the SLAM system,the optical flow method is introduced to quickly track the motion of the feature points in the image sequences and compared with the traditional feature matching method,then a combination method of optical flow and feature matching is proposed.In image registration,In order to estimate the pose of mobile robot in real time,optical flow method is used to track the motion of image feature points.In addition,in order to eliminate the accumulated error in the pose estimation of mobile robots,the feature matching method is used to increase the constraint between the robot poses.At the same time,a new strategy for local loops and random loops combination is designed to improve the efficiency of the following pose optimization.In the poses optimization,the g2 o optimization algorithm is used to optimize the robot poses according to the image registration and closed-loop detection of the pose constraints.The performance of optical flow and feature matching is analyzed by experiment,and experimental results based on publicly available datasets show that the operating efficiency is improved by 28.5%,under the premise of ensuring the positioning accuracy of the SLAM system,which effectively improves the real-time performance of the SLAM system compared with the traditional feature matching algorithm.Finally,the results of online experiments show that the combination algorithm can accurately construct the trajectory and three-dimensional map of the mobile robot in real time.