A Feasibility Study on Crack Identification Utilizing Images Taken from Camera Mounted on a Mobile Robot. (2017)

• Record Type: Journal Article
• Authors: Kim, C.W.; Chang, K.C.; Sasaka, Y.; Suzuki, Y.
• Is Part Of: Procedia engineering. Volume 188(2017)
• Page Start: 48
• Page End: 55
• Publication Date: 2017
• Subjects: bridge inspection -- crack identification -- image-based inspection -- robot sensing system; Engineering -- Congresses; Engineering -- Periodicals; Engineering; Conference proceedings; Periodicals; 620.005
• DOI: 10.1016/j.proeng.2017.04.456
• Languages: English
• ISSNs: 1877-7058

Abstract

Abstract: Many roadway and highway bridges have been suffering from aging and deterioration problems. To inspect those bridges more efficiently and accurately, the Japanese government launched a series of national projects aiming to develop various innovative robotic inspection systems to support conventional visual inspections. This study is devoted to developing the sensing modules compatible with the robotic inspection system. To preliminarily investigate the feasibility of the image-type sensing modules, a laboratory experiment was conducted, taking a commercially available digital camera and a digital video camera as the sensors and a model vehicle moving on rails as the robot. Two concrete blocks were placed at a certain distance away from the sensing system and serving as inspection targets. On still images taken by the camera, it was verified that the clear identification strongly depended on the short object distance, bright target surface, and quick shutter speed. Herein, the following condition presented a successful identification: 1-m object distance, 2300-lx illuminance, F2.8 aperture, 1/250-s shutter speed, 4608 × 3456 image resolution, and theoretical space resolution 0.09 mm/pixel. Longer object distance, faster moving speed, darker object surface and poorer theoretical space resolution would decrease the identification level. In videos taken by the digital video camera, it was verified that an object distance as short as 0.17 m could provide a high quality video from which the crack could be successfully identified. Those observations provided a useful basis for further development of the robot sensing system.