Research On Mobile Robot ORB-SLAM Based On Visual Inertia Information

The development level and application level of mobile robot technology represent the level of industrial automation technology in the country,and it has extremely important strategic significance for important aspects such as science and technology and national defense.It also reflects the level of national science and technology.If the robot is in some unknown or complex geographical environment,the key technology for the mobile robot to fully auto-navigate is Simultaneous Localization and Mapping(SLAM).With the rapid improvement of computer hardware level and the rapid development of computer vision,visual SLAM has become the mainstream research algorithm in the field of SLAM.This paper focuses on the ORB-SLAM algorithm that integrates inertial information.Mainly to study the following:(1)First,a variety of camera sensors used in visual SLAM are compared and analyzed,and a depth camera is selected as a visual SLAM sensor and calibrated.The principle of the ORB-SLAM algorithm is studied,including the tracking part of the ORB-SLAM,local mapping and closed-loop detection.(2)Secondly,the ORB(Oriented FAST and Rotated BRIEF)feature points are deeply studied,and the shortcomings of ORB feature points are analyzed.An image entropy and image pyramid fusion algorithm is proposed to solve the scale invariance of ORB features.An image force field theory based method is used to improve the descriptors of ORB feature points,and the stability of ORB features is increased.PROSAC is used in feature matching.The algorithm optimizes the ORB feature matching algorithm and greatly improves the accuracy of the ORB feature matching algorithm.(3)Thirdly,by analyzing the visual SLAM to deal with the problem of poor pose estimation of fast motion,an ORB-SLAM algorithm combining inertial information is proposed.In the algorithm,considering the difference between the data frequency of the IMU(Inertial Measurement Unit)and the camera frame rate,the pre-integration method is used to process the IMU data to ensure that the IMU data and the camera data can be data-fused;Use keyframe-based IMU auto-initialization to reduce the bias and error caused by the IMU itself.(4)Finally,the dataset experiment was carried out using the Inter Corei5-7500 processor and the PC end of 8GB RAM as the experimental environment of the SLAM system.The experimental results verified that the ORB-SLAM system with IMU data is coping with fast motion or rotation is better.