2D LiDAR and camera fusion in 3D modeling of indoor environment

Abstract

Detailed 3D modeling of indoor scene has become an important topic in many research fields. It can provide extensive information about the environment and boost various location based services, such as interactive gaming and indoor navigation. This paper presents an indoor scene construction approach using 2D line-scan LiDAR and entry-level digital camera. Both devices are mounted rigidly on a robotic servo, which sweeps vertically to cover the third dimension. Fiducial target based extrinsic calibration is applied to acquire transformation matrices between LiDAR and camera. Based on the transformation matrix, we perform registration to fuse the color images from the camera with the 3D points cloud from the LiDAR. The whole system setup has much lower cost as compared to systems using 3D LiDAR and omnidirectional camera. Using pre-calculated transformation matrices instead of feature extraction techniques such as SIFT or SURF in registration gives better fusion result and lower computational complexity. The experiments carried out in office building environment show promising results of our approach.