Research On Simultaneous Localization And Mapping Based On The Mobile Robot With RGB-D Camera

Simultaneous Localization and Mapping(SLAM)of mobile robots in unknown environment is an important research topic for the complete autonomous movement.In recent years,For the reason that rgb-d camera has more obvious advantages than previous ones used in SLAM,such as high measurement accuracy,low price,fast acquisition speed,small size and the ability to simultaneously collect depth information and color images,RGB-D SLAM is gradually becoming the focus of researchers.The current RGB-D SLAM algorithm has problems such as low efficiency,low precision,and large error.To solve above problems,this paper made several work,specific as follows:(1)Discuss the necessity and significance of this research,then analyze the classical visual SLAM framework and the existing RGB-D SLAM algorithm.(2)Describe the structure,working principle and camera calibration of the RGB-D sensor,and obtain the data needed for subsequent experiments.In the previous RGB-D SALM algorithm,there are a large number of invalid points in the collected depth data.Based on the principle of median filtering and connected domain,this paper proposes a depth map repair method to improve the efficiency of the algorithm.(3)Use the ORB for feature extraction and matching,and use the Gaussian pyramid to make the ORB features have a certain degree of scale invariance,speeding up the matching speed.(4)The traditional RANSAC algorithm is suitable for a single model and can't get the best results meet the requirements.In this paper,the ORB features were extracted for feature matching,and error matching was removed with PROSAC,which improved the accuracy of the algorithm.(5)After using the Kinect sensor to obtain 3D point cloud information,the method of combining statistical filtering with voxel filter was proposed to remove noise points and reduce the consumption of storage space.(6)Aiming at the problem of inaccurate pose estimation,the traditional ICP algorithm and PNP algorithm were combined to improve the registration success rate.Finally,the SLAM system based on RGB-D octree map was implemented-which verified the feasibility of rgb-d SLAM technology.