Reference Template Generation for Automatic Target Recognition (ATR) Systems

摘要

This paper discusses Northrop Grumman's solution to the problems associated withexisting methods for the generation of reference templates for Automatic Target Recognition (ATR) Systems. Reference templates are used by ATRs of sensor systems to provide sufficient definition of the target in order to autonomously detect, identify, and track the target. Existing methodologies typically rely on the expertise of the user to predict the characteristics of the target in the waveband of the sensor and to manually create a reference template. Such reference templates are limited by the conditions for which they were planned and by the knowledge base of the user. In addition, they do not provide any prediction of the success associated with the reference template. Northrop Grumman has developed and validated a method of automatic reference template generation for imaging infrared (IIR) sensors that eliminates these limitations and inadequacies. By utilizing a prediction of the sensed target image and the sensor's ATR, the method generates a reference template that is insensitive to changes in conditions that affect the sensed target image. The method also provides a prediction of the success of the reference template by means of a probability of acquisition (PACO). Although developed for IR sensors, the methodology is applicable to other advanced sensor systems.