Options
2000
Conference Paper
Titel
Flexible multisensor image exploitation by model based ATR
Abstract
There is a growing demand for assistance by ATR (Automatic Target Recognition) components in military image exploitation systems in order to speed up the image interpretation process and to improve the quality of the results. Besides low level exploitation tools also several high level ATR algorithms are currently tested as automatic image exploitation modules in different experimental systems. These algorithms are only useful if their performance, speed and reliability conform with the process requirements. In order to cope with the rapid progress in sensor technology, varying international mission scenarios, and target objects especially in the military domain, flexible and adaptive concepts have to be used. The necessary flexibility and expandability of ATR algorithms can be achieved by the model based approaches presented in this paper. The knowledge about the sensors, the objects, the scene and the dynamic behaviour is therefore represented by explicit models which are matched against the imagery. The additional modelling of the properties of different sensors enables sensor fusion on different levels. All these models are easily adapted to represent new objects or tasks or new sensor properties. The results of four such model based ATR algorithms of different model complexity using SAR, IR and VIS imagery are presented as two prominent examples: image registration and man-made-object screening.
Language
English