Automatic Pattern Recognition in Medium Wavelength Infrared Imagery: Part I. Search for Stable Features

Under fluctuating meteorological parameters of the atmosphere (temperature, humidity, etc.), the infrared image can dynamically change its contrast and brightness characteristics. This leads to the fact that the morphological imaging pattern of the detected object is subject to changes. The main reason is the inhomogeneity of the atmosphere, which affects the propagation of infrared radiation from the source to the infrared camera, as well as the profile of the object, which characterizes its heat distribution in the field of view of the infrared camera. In this paper, we searched for the stable features of the morphological imaging pattern for automatic pattern recognition in the MWIR (3–5 $$\mu$$ m) relative to weather conditions and the object profile. It is shown that the feature of the middle infrared contrast, which is the difference between the warmest elements of the object and the image background, is an invariant characteristic that can be used in automated target recognition problems. The results of the experiment for vehicles demonstrate the effectiveness of this parameter relative to changing temperature and humidity of the environment. However, the profile of the object can distort the ‘‘stability’’ of the middle infrared contrast due to the geometric arrangement of the heat-carrying vehicle elements.