Automatic vehicle detection in infrared imagery using a fuzzy inference-based classification system

This paper describes a unique approach of using a fuzzy inference system for target detection and classification. It first describes the methods that are used to identify regions of interest within each frame of the infrared imagery. Next, the specific data features that are extracted from these regions of interest are described. The fuzzy inference system used in this application is described. This description includes discussions of the feature input and system output membership functions, the rules used in the inference system, and the logical operations, implication, aggregation and defuzzification methods employed. Finally, results attained by applying the described approach to a "blind" closing sequence data set are provided and conclusions are drawn. The developed techniques have proved to be robust and have demonstrated an ability to properly classify a variety of targets in different clutter environments. The described approach can easily be expanded to utilize other feature inputs.

I. Introduction

INFRARED sensors have been used in missiles and other systems for the detection and classification of tanks and other vehicles. The sensors exploit a combination of temperature differences, emissivity differences, and “cold sky” reflections that in combination result in imagery with a high contrast between the tank or other vehicle target and background clutter. In many cases this contrast is superior to that which would be attained in visible imagery. Closing sequence infrared imagery refers to image data collected using infrared sensors mounted on a missile. The closing sequence is a series of infrared images that are collected as the missile approaches and eventually encounters its intended target. For some years it has been desired to automatically detect and classify tanks and other vehicles in closing sequence data with an overall desire to develop air to ground missile systems with a fire and forget capability.