Proceedings of the
9th International Conference of Asian Society for Precision Engineering and Nanotechnology (ASPEN2022)
15 – 18 November 2022, Singapore

LIDAR-based Automated 3D Inspection System for Corrugated Surface Defects

Chaoyu Dong1,2, Zhaoheng Shi1 and Hong Luo1,a

1Singapore Institute of Manufacturing Technology, A*STAR, 2 Fusionopolis Way, 138634, Singapore

2School of Computer Science and Engineering, Nanyang Technological University, 50 Nanyang Ave, 639798, Singapore


With the advantages of high stiffness and long durability, corrugated surface structure is widely employed in the manufacturing, construction, and transportation areas. Most corrugated surfaces are deployed outdoors and exposed to external crushes and weather changes. To improve operating performance and reduce maintenance efforts, an effective and efficient defect inspection system is of great importance. However, the varying size and shape of corrugated surfaces challenge the human inspection process leading to low flexibility and efficiency. Besides, uncontrolled environmental lighting and weather condition burden the adoption of traditional imaging cameras. To address the demanding inspection requirement for corrugated surface defects, this work develops a laser imaging, detection, and ranging (LIDAR)-based automated 3D inspection system. The system is composed of an online LIDAR scanning platform and defect inspection algorithm. By virtue of infrared light pulses in LIDAR, the scanning process and results are not affected by environmental lighting variations. The scanning platform is designed to capture only one shot for the surface's short edge and continues scanning along the long direction. Under this scanning mode, the wide field of view and near filed accuracy of LIDAR are fully utilized. The fine and densely grained morphology information along the surface's short edge can be rapidly formulated in the 3D point-cloud view. The complete 3D point cloud is then automatically analyzed by a four--sequential stage algorithm: cloud transformation, plane simulation, abnormality clustering, location and area determination. The point cloud range is filtered and calibrated to identify and transform the corrugated surface with its back bottom corner as the origin. Through the iteration of random sample consensus, an approximate plane is simulated to segment normal and abnormal points. The Euclidean distance is adopted for abnormality clustering to mesh abnormal points according to their neighbouring intervals. 3D bounding boxes are generated for the abnormal clusters. The centroid and area of the bounding boxes are found to quantify the defect location and dimension. The proposed LIDAR-based defect inspection system and algorithm have been evaluated in the inspection of defects like dents and deformation for manufacturing corrugated surface structures.

Keywords: LIDAR, 3D automation, Defect inspection, Corrugated surface

PDF Download