Real Time Condition Monitoring Using Camera and Phase-Based Motion Estimation

Abstract

Defects on manufacturing equipment, such as abrasion, corrosion, and deposition, may decrease the production quality and even lead to the shutdown of the entire manufacturing line. Various condition monitoring technologies have been developed to predict machine health and safety through contact sensors. The noncontact camera has shown many advantages over the conventional contact sensors, including high spatial resolution, low cost and remote sensing. Moreover, phase-based motion amplification has proven to be an efficient tool for detecting subtle vibrations. However, the motion amplification requires significant computational resources and does not provide direct motion signal output. To detect abnormal vibrationsduring long-term inspections, a more efficient phase-based motion estimation technique is necessary. In this study, we propose a real-time vibration monitoring system that uses a camera and image phase. We use single optimal Gabor filter with phase-based optical flow to extract the vibrational motion. The use of single optimal filter significantly reduces computation costs and enables accurate measurement of vibration signals even in the low-light condition and/or with image noise. Parallel computing scheme is also introduced for real-time condition monitoring. We conducted validation experiments on a structure with multiple vibrating components that simulate a real factory line. Damage detection is performed on the structure with two damaged cases. All the results show that the proposed technique can accurately measure displacement and provide a novel solution for camera-based real-time damage detection.