Automatic Target Recognition by Infrared and Visible Image Matching

Automatic target recognition based on the long wave infrared (LIR) and visible-light spectrum (VIS) image matching is a very challenging problem. It is due to the fact that the images between LIR and VIS have lots of different textures. This difficulty is inherent in the thermal radiation imaging affected by one of the principal mechanisms so called heat transfer. In this paper, a novel algorithm is presented for object recognition between the LIR and VIS images under various conditions. It is assumed that the visible light images of the target are available a priori, and the newly acquired infrared images are used to perform the target recognition task. The LIR and VIS images are first initialized with edge detection and binary template matching, followed by a local fuzzy threshold to identify the high similarity objects. Our method has has low computational requirements and can be implemented on a real-time system. Several experiments are carried out using the real scene images with various test objects.