A Laser Data Compensation Algorithm Based on Indoor Depth Map Enhancement

Abstract

The field of mobile robotics has seen significant growth regarding the use of indoor laser mapping technology, but most two-dimensional Light Detection and Ranging (2D LiDAR) can only scan a plane of fixed height, and it is difficult to obtain the information of objects below the fixed height, so inaccurate environmental mapping and navigation mis-collision problems easily occur. Although three-dimensional (3D) LiDAR is also gradually applied, it is less used in indoor mapping because it is more expensive and requires a large amount of memory and computation. Therefore, a laser data compensation algorithm based on indoor depth map enhancement is proposed in this paper. Firstly, the depth map acquired by the depth camera is removed and smoothed by bilateral filters to achieve the enhancement of depth map data, and the multi-layer projection transformation is performed to reduce the dimension to compress it into pseudo-laser data. Secondly, the pseudo-laser data are used to remap the laser data according to the positional relationship between the two sensors and the obstacle. Finally, the fused laser data are added to the simultaneous localization and mapping (SLAM) front-end matching to achieve multi-level data fusion. The performance of the multi-sensor fusion before and after is compared with that of the existing fusion scheme via simulation and in kind. The experimental results show that the fusion algorithm can achieve a more comprehensive perception of environmental information and effectively improve the accuracy of map building.