Research And Application Of Laser SLAM Algorithm Based On Visual Line Features

Instant localization and mapping(SLAM)is a key technology that enables mobile robots to perceive their own location and achieve safe navigation.Due to the increasing complexity of the environment faced by robots,it is of great significance to study how to extract higher-level environmental features from multi-source sensor information,and to implement a multisensor fusion SLAM algorithm that can effectively operate,in order to promote the development of robot intelligence.In order to enable robots to better understand the environment and improve the positioning accuracy and closedloop performance of two-dimensional laser SLAM algorithm.this paper designs and implements a laser SLAM algorithm that integrates visual line features.The main research contents include:(1)Summarize the current situation of SLAM fusion between vision and LiDAR,analyze the measurement principles of sensors,establish mathematical models of each sensor,and complete the identification of model parameters through calibration experiments,providing a theoretical basis for subsequent pose fusion.(2)We have constructed a visual subsystem that provides line feature pose constraints.To address the problem of limited perspective and poor performance of the line feature plane construction method for mobile robots,the endpoint method is used to construct spatial straight lines,and a weighted line fitting association algorithm is proposed to obtain the initial depth of line features from depth maps.By establishing a line feature re projection error model and a sliding window optimization strategy,good visual constraints are provided for backend pose fusion.(3)A laser fusion localization mapping algorithm incorporating visual line features has been designed and established.To solve the problem of nonlinear dimensional degradation of sensors,a degradation factor is introduced,and only updates in the non degradation direction are carried out during optimization.At the backend of the system,a graph optimization framework for pose fusion has been established,which effectively improves the positioning accuracy and mapping effect of the algorithm in this paper by minimizing the joint error between vision and laser.(4)In response to the problems of large cumulative errors and poor closed-loop performance of the two-dimensional laser SLAM algorithm in degraded and dynamically changing environments,a weighted line fitting algorithm is proposed to extract laser line features and construct a submap based on line features.By determining whether there are geometric and appearance consistent features between the line feature submaps,the closed-loop detection ability of the SLAM algorithm is strengthened.(5)Conduct comprehensive testing and evaluation of the algorithm’s positioning accuracy,mapping effect,and closed-loop capability in various complex environments on the OpenLORIS dataset.The results show that the positioning accuracy and closedloop detection ability of the proposed algorithm have been effectively improved in complex environments.Finally,the effectiveness and practicality of the algorithm were verified in real application scenarios by building a robot experimental platform.