Automatic Detection of the Interiorly Welded Joints of Pipelines from Video for Remote Visual Inspector

For sake of reliability, remote visual inspections of the quality of the interiorly welded joints of the city pipelines are still performed in cities such as Kunming of China from the videos which are captured and wirelessly transmitted by a small pipeline vehicle when the pipeline construction project undergoes the acceptance-checking. It is tedious and tiring for the inspector to visually finding out the welded joints from the videos continuously captured and wirelessly transmitted by the pipeline vehicle. In this paper, we propose a method for automatically detecting the interiorly welded joints to reduce observing load of the inspector. Our method consists of filtering frames that possibly have no the welded joint contours, filtering the horizontal band electromagnetic noises that cannot be filtered with conventional noise filters, detecting the contour of the welded circular joint by accelerated Hough transform and estimation the orientation angle of the joint section of pipeline relative to the moving direction of the vehicle. Experiments show that our method is effective and satisfying real-time requirement with respect to the moving velocity of the vehicle and the time interval of capturing and transmitting a frame.