Discrimination of targets from background of similar temperature, using two-channel data in the 35–41-μm and 11–12-μm regions

Abstract

Target detection for thermal IR imagery is usually accomplished by pixel intensity thresholding routines. Hot targets (vehicles with engines running) appear as high-intensity (bright) areas on IR imagery. A threshold-limited detector shows all pixels above a specified intensity value as targets that have, at the time of imaging, a large temperature differential from the background. However, when a significant temperature differential is not present, targets are very difficult to detect by simple thresholding. This study has addressed the problem of detecting targets which are close to ambient temperature. This was done by modeling the background and by using the model as a base for developing an understanding of physical processes and their variations, which will (with further work) lead to the construction of a filtering algorithm to enhance target/background contrast. We have shown that by ratioing radiance received from the target and from the background, better target discrimination is possible than is the case from using one bandpass, especially when target and background are of a similar temperature.