Automatic alignment of infrared video frames for equipment leak detection

Abstract

Petroleum refining and petrochemical industries start using infrared (IR) cameras to detect volatile organic compounds (VOC) leaking out of process equipment. However, further quantitative processing of these video image data or automatic recognition of VOC plumes are hindered by unaligned video frames owing to the slight vibrations of the camera. An automatic method is developed to align the IR video frames as a preprocessing procedure for other possible video processing methods. The alignment method is based on a two-dimensional spatial Fourier transform. The accuracy can reach fractional pixels in estimation of translational shift and 1–2° for rotational shift. Temporal Fourier transform of actual industrial tests of IR videos is performed with both unaligned and aligned video frames. The results indicate that only after the alignment of the video frames, the camera motion interferences on VOC plume identification can be eliminated or minimized, and the VOC plume can be identified through investigating the characteristic flickering frequency power in the temporal Fourier transform. This alignment method provides a useful tool for IR or other optical video image data preprocessing purposes.