Research On Visual-inertial SLAM System Based On Point And Line Features

Simultaneous Localization and Mapping is the core technology in the field of computer vision and robot navigation.SLAM technology is widely used in the field of robots.Its purpose is that robots can estimate their position in real time based on sensor data without the prior knowledge,and at the same time build an environment map.The SLAM system based on point features has recently become the focus of people's research,but in a structured environment,line features are as rich as point features.This paper proposes a binocular visual inertial SLAM system based on point and line features,the purpose is to fuse point features,lines Features and IMU information improve robot positioning accuracy and robustness.The main research contents of this article are as follows:Firstly,this article builds a binocular IMU visual inertial module.According to the realtime and synchronization requirements of the SLAM hardware platform,the module is introduced from the component selection,circuit design to the overall framework,and the FPGA driver design and IMU data acquisition driver design of the binocular IMU visual inertial module are carried out Introduction,and then perform module performance testing.Secondly,this paper studies the extraction,matching and parameterization of line features.It introduces in detail the principle of using LSD method to extract line segments in the image and the method of generating LBD descriptors,and then carefully analyzes the principle,advantages and disadvantages and application scenarios of the Pluck representation method and orthogonal representation method of spatial straight lines.Then,this paper proposes a visual inertial tight coupling optimization algorithm based on point-line features.Based on the theory of sliding window optimization,a method of simultaneously combining binocular visual constraints(point features and line features)and IMU constraints is proposed.The optimization objective functions are priori residual,IMU residual,point feature observation residual and line The feature observation residuals are deduced in detail.At the same time,this paper proposes a key frame selection mechanism,and analyzes the sliding window optimization algorithm and the principle of marginalization.Finally,this paper builds a binocular visual inertial SLAM system based on point-line features for experiments,including hardware platform construction and software system.Experimental comparison with several other mainstream algorithms on the Eu Ro C data set verifies that the binocular visual inertial SLAM system based on point and line features proposed in this paper has high positioning accuracy and robustness,and experiments are conducted in outdoor scenarios,Verified that the hardware platform built in this paper has high stability and reliability.