Research On Simultaneous Localization And Mapping Based On Computer Vision

Intelligent manufacturing is the direction and trend of rapid development of today's manufacturing industry after the Industry 4.0.It has great applications in the military field,life service field and industrial Internet of Things.Among them,industrial intelligent robots will be an important component of intelligent manufacturing,replacing human beings to complete repetitive or dangerous work.The intelligent robots can quickly and efficiently complete production tasks in industrial production.Some industrial intelligent robots need to be able to judge their outer environment and analyze it.After that it need to design the path in the process of completing tasks.Therefore,the research on the autonomous localization and mapping algorithm of industrial intelligent robots is particularly important.This thesis researches on autonomous localization and mapping algorithm,and proposes a complete autonomous localization and mapping system.After that,this paper compares and analyzes system's main algorithms.This paper deeply explores the model of the vision sensor and the calibration correction process,and compares and analyzes the interest point extraction and interest point matching algorithms in the front-end visual odometer.Finally,the real-time and overall accuracy of the overall localization and construction are demonstrated and analyzed through experiments.The general thoughts and main work of this paper are as follows:In order to establish a complete robotic autonomous localization and mapping system,and make system platform solution combining software and hardware,this thesis firstly selects the appropriate visual sensor and robot mobile platform as the hardware foundation of the system establishment.Then it uses ROS(Robot Operating System)system to realize communication and data transmission between modules in the system,and finally using Simultaneous Localization and Mapping algorithm to achieve autonomous localization and mapping functions.The imaging model of the vision sensor used in the system was analyzed to reduce the error of the image acquired during the imaging process.Analyze the visual sensor calibration algorithm,which is to adjust the basic parameters of the visual sensor and various distortions correction methods to improve the accuracy of the image,and analyze and demonstrate the algorithm by experiment.Theoretical analysis and experimental verification of the front-end visualmileage calculation method in SLAM algorithm are taken.The front-end visual odometer includes three processes of interest point extraction,interest point matching,and camera pose estimation.The real-time and accuracy of the three algorithms in interest points extraction are analyzed and compared with the combination algorithms of three interest points matching algorithms.Among them,SIFT algorithm's feature extraction time is about 0.4ms,SURF algorithm's feature extraction time is about 0.12 ms,and ORB algorithm's feature extraction time is about 0.02 ms.Therefore,ORB algorithm's real-time performance is better than the other two algorithms.The feature extraction and feature matching algorithms with high processing speed and high precision are obtained.And the conclusion is verified by experiments.Finally,through the experiments and analysis of the overall autonomous localization and mapping system,the hardware system construction,the communication of each module of ROS system and the realization of autonomous localization and mapping algorithm are completed.The preparation of the whole experiment and the robot mobile platform and the construction of the experimental system are completed.Through real-time analysis of the system,the system processes image frames around 25 frames per second,which can meet the real-time requirements of localization and mapping.Analyze the absolute error and relative pose error of the system to evaluate the accuracy of the trajectory,analyze and demonstrate the algorithm by experiment.