Research On Localization And Mapping Method Based On RGB-D Camera

SLAM technology is the basis and prerequisite for autonomous navigation of robots,which has been widely used in the field of robots.However,the Global Navigation Satellite System(GNSS)becomes inaccurate in the indoor environment.Therefore,it is necessary for robot localization and mapping using its own sensors.Because the RGB-D camera is inexpensive,portable,and capable of providing both RGB image and depth information,it has attracted great attention recently.In this paper,we studied on the localization and mapping method based on RGB-D camera.The main work included as follows:1.In most RGB-D SLAM system,the RANSAC framework was applied to eliminate false matching points in the feature matching part.However,it is time-consuming.This paper presented a feature point matching method based on random four-point order-preserving constraint,and this method improved the real-time performance of the system.The matching experiments using public datasets validated that the proposed method was effective and feasible.2.In order to solve the problem that the pose estimation of RGB-D SLAM system may be inaccurate or even lost in the scenes with few textures,this paper presented a strategy to fuse IMU data and RGB-D information.This proposed SLAM system become more robust and can continue working stably for indoor environment with sparse textures.3.The monocular SLAM system based on the fisheye camera was given in this paper.A large number of experiments with fisheye images and fisheye videos verified the feasibility of this method.The experimental results have shown that it is also the effective supplement and extension for the RGB-D SLAM method.