Enhancing Long-Range Automatic Target Recognition Using Spatial Context

This paper presents a high-performing automatic target recognition system which can be used for long-range surveillance scenarios. The main novelty of our system is that it uses contextual information from RGB images to help classify targets in long range real world LWIR images. This contextual framework provides additional information of an object's surrounding environment, leading to a significant increase in long-range target recognition accuracy. This work will be of interest to the defence community as a high-performing automatic recognition system is a highly sought-after capability.