RGB-D SLAM Method Based on Object Detection and K-Means

Abstract

Aiming at the problem that the traditional visual simultaneous localization and mapping (SLAM) algorithm is easily affected by moving targets in dynamic environment, which leads to the degradation of system localization accuracy, a visual SLAM algorithm based on object detection and K-Means is proposed for application in dynamic environment. It incorporates the YOLOv5n object detection network with the addition of a leak detection judgment and repair algorithm and the K-means clustering algorithm, which effectively rejects dynamic objects in images and maximizes the retention of static information. Experiments on publicly available datasets show that the error of this paper's method is smaller than that of other SLAM algorithms applied in dynamic environments, and it can guarantee real-time operation.