Multi-source fusion SLAM system based on visual point-line feature optimization

Abstract

The invention discloses a multi-source fusion SLAM (Simultaneous Localization and Mapping) system based on visual point and line feature optimization. Firstly, a visual line feature extraction method of an improved scale space is provided, and constraint matching is performed on the same line feature between a front frame and a rear frame by using a constraint matching strategy based on a least two norm, so that richer feature information is provided for the front end of a visual odometer. Secondly, the multiple frames of laser point clouds is projected into a visual coordinate system to realize depth correlation between the laser point clouds and visual features, and the point cloud is used for assisting in optimizing laser radar scanning matching precision by using a visual initial pose estimation result; and finally, a laser-vision-inertial odometer system is constructed by adopting a factor graph method based on a Bayesian network, and a GNSS factor and a loop factor are introduced to carry out global constraint on the laser-vision-inertial odometer. According to experimental comparison, the algorithm is superior to similar algorithms in the aspects of real-time performance, positioning precision and mapping effect, and real-time pose estimation with excellent positioning and mapping precision can be realized in an EuROC data set.

Description

technical field [0001] The invention belongs to the field of multi-sensor real-time positioning and mapping schemes, and in particular relates to a multi-source fusion SLAM system based on visual point-line feature optimization. Background technique [0002] The multi-source fusion positioning technology based on SLAM (Simultaneous Localization and Mapping, real-time positioning and mapping) is one of the key technologies in the field of high-precision positioning of mobile carriers. According to the different sensors, it can be divided into two types: laser SLAM and visual SLAM. Due to the inherent defects of a single sensor such as the limited scanning angle of the lidar and the significant impact of the visual odometer on illumination changes, in recent years, the lidar-visual-inertial mileage LiDAR-Visual-Inertial Odometry (LVIO) has become a research hotspot in SLAM due to its multi-sensor heterogeneous complementary advantages. [0003] The existing LVIO multi-source ...

AI Technical Summary

Problems solved by technology

However, the extended Kalman filter method is used to loosely couple the GPS positioning information with the lidar pose estimation value, but there is a disadvantage of a single linearization error.

The LVIO framework based on the factor graph method that integrates GNSS factors can optimize the pose estimation value through multiple iterations of the sliding window, but the measurement information of a single key frame in the factor graph is too much, resulting in redundant information and high computational costs.

Images