Automatic Aerial Target Recognition On Infrared Imagery

With the help of advances in imaging technologies, cameras working at different spectral bands started to be used in a wide area of applications including industrial, civil and military fields to provide novel solutions. Infrared imaging becomes one of the examples of such novel technologies, due to the fact that infrared images and visual images have different characteristics. Recently, the methods that are used for feature extraction and classification in visible images are also experimented on IR images and their performances are also investigated in studies. In this work, histogram of gradients (HOG), moment and convolutinal neural network are applied to the problem of automatic target recognition using infrared imagery. Performances of the methods are investigated on synthetic and real IR images.