Indoor 3D reconstruction and mapping for high ceiling spray painting robot

Sep 2016 - May 2017 Linhao Jin

Recent robotics advancements allow researchers to boost productivity in construction industry. Two famous construction robots, Quicabot and Pictobot, are designed for the inspection and high rise spray painting purposes for newly finished buildings.

This webpage present a perception strategy that is used to preform 3D reconstruction for the unknown environment. This strategy is adopted by the high rise painting robot, Pictobot, for planning and localization purposes. The 3D reconstruction is achieved using a sensing system comprising a tilting laser scanner, SICK LMS100 fixed on the Oriental Motor AZ46AK-PS36 stepper motor. The sensing system is carried by a mobile robot platform.

General flowchart

Scanning and tilting strategies presented in this work aim at synergistic incorporation of obtaining and merging 3D point cloud from sensor outputs. The stepper motor moves 1 degree per steps and covers the viewing angle of 200 degrees in the end. After scanning operation, all point clouds obtained are registered using Normal Distribution Transformation (NDT) algorithm.

Software architecture

After statstical outlier removal, the final mapping output is shown below. Red and blue point clouds represent two 3D models from different views.

Mapping output

Codes available at:
Multi-layer Insultation segmentation using FCN
Stereo Vision-based Gantry Detection for Autonomous Driving
Building 3D Texture Model from Video and CT for Endoscopic Sinus Surgery