Wavelength diversity based infrared super-resolution and condition-based maintenance

The relevance of image sequence super-resolution (SR or high resolution (HR)) and wavelength diversity to encompass IR imaging for condition-based maintenance (CBM) of complex systems was pointed out last June at the previous Forum at Harrogate in 2007. In CBM, non-contact analysis of plant condition to faulty components (fuses, bearings, motors, steam pipes, gears, belts) can prevent costly shutdowns by corrective maintenance and breakdown prediction to enable pre-emptive action. Since HR infrared cameras are very expensive, digital super-resolution techniques can be used to create an HR image from low cost low-resolution (LR) images. Wavelength diversity spans the range from ultraviolet (0.1 microns) to infrared (1100 microns). Thermal IR (infrared) images are focal plane IR detector (bolometer) array and lens combinations that give a visual representation of IR energy emitted. Thermal IR images can passively see all objects regardless of ambient light and thermography can give detailed images of invisible objects. IR thermography is safe, non-intrusive and efficient for identifying faults and overheating components.

Images gathered by an array of IR detectors (staring array, microscanning array) are inherently of limited resolution because of the constraints on the size of the pixels and geometry of the array. Trade-off between energy gathering capability and noise is required, especially because of the particular characteristics of IR image sequence acquisition including the presence of significant aliasing. This paper presents the super-resolution technique based on moving least squares for IR image sequence super-resolution. Results of super-resolution on simulated sets of undersampled degraded IR images are also presented. A procedure for higher-order whitening in the transmitting end and de-whitening in the receiving end is also considered within the overall framework of thermal image data acquisition, compression, coding and transmission of the data through communication channels followed by de-whitening and image sequence super-resolution at the receiving end.