다시점 이차원 영상처리 기술을 활용한 화력발전소 보일러튜브 감시 시스템

초록

This paper proposes a monitoring system that can measure the volume of the attachments on boiler tubes. For 3D visualization, the system uses a combination of stereo matching and 2D stitching algorithm to check internal environment of the boiler in thermal power plant. Based on the estimated volume of the stereo algorithm, quadratic fitting model is used to correct the predicted volume of attachment. The volume measure algorithm achieves mAP of 87.08% on ICVL ICE dataset and the 3D visualization also helpful for monitoring the growth of attachments.