Research On SLAM Of Indoor Mobile Robot Based On RGB-D Camera

Simultaneous Localization and Mapping(SLAM)is the key technology for mobile robots to achieve truly autonomous motion.In recent years,the SLAM algorithm using RGB-D cameras as sensors has become a research hotspot in the field of robotics.At present,the RGB-D SLAM algorithm has made great progress,but there are still the following problems:(1)The feature point matching accuracy is low and the efficiency is not high;(2)single visual odometry algorithm is not adaptable to various scenes;(3)loop closing efficiency and accuracy are low.Aiming at the problem of low accuracy and inefficiency of feature point matching,this thesis proposes an improved feature point matching algorithm,which filters the randomly sampled samples before the feature points are matched to eliminate the obviously wrong match point pair.The algorithm can reduce the number of calculations of RANSAC while improving the correct rate of the matching algorithm,thereby reducing the time of feature matching.The matching experiments on the TUM dataset show that the improved algorithm can effectively improve the accuracy and efficiency of feature matching.Aiming at the problem that the single visual odometry algorithm is not adaptable to various scenes,this thesis proposes an improved visual odometry algorithm and proposes the selection strategy of the algorithm.According to different scenes,this selection strategy can automatically switch to improve the adaptability of the algorithm to the environment.Camera tracking experiments on a large number of TUM datasets show that the algorithm is more adaptable to various scenarios than single visual odometry algorithm.Aiming at the problem that the loop closing algorithm for BoW model is easily limited by the size of visual dictionary,resulting in low detection efficiency and accuracy.This thesis proposes an improved loop closing algorithm,which uses the loop closing mode from key frame to local map.The visual dictionary tree is introduced to describe the scene,which speeds up the search speed of the feature,thereby improving the efficiency of loop closing.At the same time,the method of confirming the true and false loops is introduced,thereby improving the accuracy of loop closing.The loop closing experiment results on the TUM dataset show that the improved algorithm can effectively improve the efficiency and accuracy of loop closing.All of the algorithms in this thesis were tested on the TUM public dataset.For the improved algorithm of feature matching and loop closing,the experiment is compared with the traditional algorithm in terms of real-time and robustness.For the improved visual odometry algorithm,the experimental aspects of the adaptability of various scenarios are mainly verified.Finally,from the overall performance of the algorithm,the proposed algorithm is superior to the traditional RGB-D SLAM algorithm.