Grid map representation with height information based on multi-view 3D reconstruction technology

This paper presents an innovative grid map representation with height information (GMRH) method, which leverages multi-view 3D reconstruction technology (MV3D) in unmanned factory settings. The primary objective is to enhance the mobile robot's capabilities in hazard identification and its comprehensive understanding of the environment. The approach presented in this paper, demonstrates significant strengths, particularly in terms of high geometric accuracy, a broad field of view, and cost-effectiveness, when addressing this problem. The solution is primarily divided into two key phases. Firstly, the MV3D method captures panoramic environmental data within unmanned factories in the form of dense point clouds. Subsequently, these dense point clouds are mapped to various planes based on the robot's safety height and presented to the mobile robot using the GMRH method. Furthermore, environmental factors such as variations in lighting, occlusions, and transparent obstacles frequently impact the accuracy of map generation. To address this issue, this paper conducts experiments on the response of reconstruction results to changes in camera position and attitude, deriving a generalized law governing their influence on reconstruction outcomes.