Intelligent Construction Machinery SLAM with Stereo Vision and Inertia Fusion

Positioning technology is the foundation of intelligent construction machinery, the current mainstream positioning solution is simultaneous localization and mapping (SLAM) technology, which is mainly divided into lidar SLAM and visual SLAM. Due to the high cost of lidar, it is easy to degrade or even fail in scenes with a single environmental texture; while the cost of vision sensors is low and has a wealth of environmental texture information acquisition capabilities, which can effectively avoid degradation problems. In order to reduce the localization cost of intelligent construction machinery and improve the positioning accuracy, based on the VINS-Fusion stereo visual-inertial tightly coupled system framework, an improved Random Sampling Consensus (RANSAC) algorithm is used to reduce feature mismatch, and the Huber kernel function is used to IMU residuals and visual residuals are constrained to improve the effect of the SLAM system. Compared with the mainstream VINS-Fusion algorithm, the positioning root mean square error of this method on the EuRoC dataset is reduced by an average of 12.41%, which improves the positioning accuracy; simultaneously, the experimental results in the actual scene show that the motion trajectory of the algorithm, it is closer to the real trajectory than VINS-Fusion, which verifies the effectiveness of the method.