Region of interest identification, feature extraction, and information fusion in a forward looking infrared sensor used in landmine detection

Abstract

A forward looking infrared sensor used in conjunction with a vehicular based landmine detection system is described. The paper will first describe the requirements for vehicularly deployed forward looking infrared sensors. This will be followed by descriptions of the basic forward looking infrared sensor, its components and its integration to the vehicular platform. This will be followed by a description of a novel method that is used to define regions of interest within the infrared images in real time. Next the specific features that are extracted from the infrared images that are used in target classification are provided. The paper will next describe a fuzzy inference system that evaluates the extracted data features and generates a mine confidence value that is used for final declaration of potential targets as mines or clutter. Specific examples of attained performance both with and without the fuzzy inference system will be provided to demonstrate the efficacy of the fuzzy inference system in evaluating data features and performing the classification. Lastly, the paper will describe other areas where the described methods are being successfully applied.