Outdoor Scene Localization And Mapping Of SLAM Based On Visual Laser Inertial Fusion

Simultaneous localization and map construction(SLAM)is one of the methods to realize robot self-localization and 3D reconstruction.The 3D reconstructed scene provides the priori of autonomous decision making for the robot.Compared with traditional 3D reconstruction methods,SLAM has the advantage of realizing incremental real-time reconstruction while positioning.Therefore,with the continuous development of artificial intelligence and robotics industry,the study of SLAM has gradually attracted the attention of researchers.The existing SLAM technology is mainly divided into two categories: Lidar and vision.Since the information obtained by the vision sensor is two-dimensional plane,when dense three-dimensional reconstruction is carried out by traditional monocular visual SLAM,the three-dimensional coordinate points need to be recovered point by point triangulation,which will consume a lot of time and computing resources.Since the laser radar sensor cannot obtain the color feature of the landmark point,the 3D model reconstructed by laser SLAM is usually lack of color information.However,compared with the visual sensor,the Lidar can accurately obtain the three-dimensional signpost points,thus avoiding the tedious process of visual triangulation and depth recovery.There is usually a certain degree of error in the operation of SLAM system,and this error will gradually increase with the accumulation of time and movement distance,especially in large outdoor scenes.This paper aims to integrate the advantages of the two sensors,vision and Lidar,to achieve more reliable positioning and mapping effects.At the same time,inertial navigation sensors are integrated to make SLAM play a stronger robust role in outdoor scenes.Finally,the algorithm is named VL-FAST-SLAM.The main contents of this paper are as follows:(1)Based on the laser inertial odometer(LIO),this algorithm integrates GPS and visual factor constraints to realize the complementary advantages of multiple sensors.The tedious triangulation process of two-dimensional vision is avoided,and the visual point features are fused with laser point features to achieve RGB three-dimensional reconstruction,so that the monochromatic three-dimensional model has more visual information.(2)Global Positioning System(GPS)constraints were introduced in the outdoor scene to limit the cumulative errors within a certain range,ensuring the stability of SLAM outdoor long-term missions.By introducing loopback detection,the system’s cumulative errors can be found and eliminated timely and effectively.(3)Through experiments on multiple public data sets,the final results show that the proposed method has accurate positioning and efficient reconstruction of dense three-dimensional point clouds,and can be run in real time on platforms such as Intel.(4)The mobile terminal Internet APP was developed to monitor the localization of SLAM algorithm and three-dimensional point cloud reconstruction in real time on the mobile terminal.(5)The 3D point cloud grid reconstructed by the algorithm is converted into a 3D grid map through open source meshing algorithm library for industrial application.