IMU-assisted robotic structured light sensing with featureless registration under uncertainties for pipeline inspection.

Laser profilometry and structured light sensors are being increasingly deployed for pipeline inspection as they provide the operator with a precise 3D map that can enable visual detection and direct insight into the integrity of the pipe. The focus of the presented paper is the design of an integrated robotic structured light sensing system used to improve the performance of 3D defect reconstruction for pipeline inspection while accommodating the uncertainty seen in a real-world environment. Point cloud registration of the consecutive 3D frames is a key factor in building this 3D map; therefore, a comprehensive featureless registration approach is proposed first, which is proven more efficient than conventional feature-based registration algorithms. Wheel odometry from the developed robotic platform and inertial measurements are integrated into the registration algorithm to enhance the 3D reconstruction performance for sensor stabilization. An intensity-based threshold searching method is further applied to retrieve the reconstructed defect size. Lastly, the uncertainties of the structured light sensing are investigated for the total reconstruction uncertainty and estimated measurement uncertainty to be quantified in order to illustrate the measurement precision. The efficacy of the proposed algorithms are supported by experimental results of pipeline inspection. • 3D profiling obtained from structured light sensors is widely applied for pipeline inspection. • Inertial Measurement Unit and Wheel odometry provide orientation and location of the robot as external input. • The proposed sensing system take the global and local positioning information from the sensor to construct a comprehensive 3D point cloud registration approach. • The proposed registration algorithm can reconstruct pipe even with no visual features in inspection, and performs better than other stats of art methods. • Uncertainty sources in this robotic integration sensing system are clarified and the uncertainty from measurement has been investigated to illustrate the reliability of the designed inspection system. [ABSTRACT FROM AUTHOR]