Detection of a target at a long range in infra-red band

Abstract

IR(Infra-Red) band is widely used in remote sensing for it is a good compromise between Microwave and Visual band. With the merit, IR band is suitable for the detection of a target at a long range, which is several km order. At such a long range, a target is usually viewed as a point with some relative brightness against its background. Due to the randomness of that relative brightness, a perfect correctness of detection can not be guaranteed. Only a probabilistic correctness called as the detectability can be asserted as a performance measure for a detector in a given condition. To predict and compare the performances of detectors in various conditions, some quantification of their performances is needed. In this thesis, an integrated framework is proposed to quantify the detectability of a IR detector when detecting an air fighter in the open air. It includes the IR signature modeling for the target, and organized steps to predict the detection probability from its signature in various conditions. A multi-color detector can be used to enhance the delectability instead of uni-color detector. Detectability measures for multi-color are also presented by extending the concept of the detection to the classification. They are the Mahalanobis distance and the Bhattacharrya distance, which are usual measures for classification. With these detectability measures, we can predict and compare the performances of IR detector for various conditions.