XCS for robust automatic target recognition

Abstract

A primary strength of the XCS approach is its ability to create maximally accurate general rules. In automatic target recognition (ATR) there is a need for robust performance beyond so-called standard operating conditions (SOCs, those conditions for which training data is available) to extended operating conditions (EOCs, conditions of known targets that cannot be foreseen and trained for). EOCs include things like vehicle-specific variations, environmental effects (mud, etc.), unanticipated viewing angles, and articulation of components of the target (hatches, turrets, etc.). This paper presents experiments where XCS addresses structural generalization over global and local features normally used in ATR classification. In many SOCs, these features are adequate for target recognition. Our goal with XCS is to form generalized rules that utilize these features for effective ATR in EOCs. Results show that XCS is effective in this generalization task. Conclusions and future directions for research are discussed.

Shape-based recognition of targets in synthetic aperture radar images using elliptical Fourier descriptors

“…The SVM classification schemes were implemented using a freely available package for Matlab™ called the Support Vector Machine Toolbox Version 0.5 23 . In general, it was difficult to compare the EFD-based classification results with the ones of previous studies, as many of them used subsets of the available images in the MSTAR dataset for both 3-target and 10-target ATR problems 11,12,24 . Some studies considered the pose (azimuth) to be known a priori, which significantly simplified the ATR problem (see for example Cetin 9 and others 11 ) and only a few did otherwise 3,5,10,25 .…”

“…Still, the use of high level features was shown to be a useful tool in ATR, considering that the smallest number of features used in the template approaches was 2304 (48×48 pixels) 3 as opposed to the four in the high level feature approach 11 . Another common feature used specific to radar is the location and magnitude of peaks 11,12,13,14,15 . In addition to the aforementioned heuristic features, additional ones, common to general image processing, are also used in ATR systems.…”

Developing a robust integrated learning system for the modern battlefield

“…Recall that robust methods are one aspect of our vision for an integrated learning system. More extensive and detailed discussion of this application is available in [16] [16]. In our work we have used the freely available XCS software from the Illinois Genetic Algorithms Laboratory [1] and tailored it to our particular application.…”

ATR Using Multi-view Morphing

“…Hence, one of the major tasks in image analysis is to identify the different objects present in the scene. Another major difficulty arises due to the fact that the operating conditions may differ significantly from those of training and are not anticipated [24]. The major issues and needs of object recognition include good representation of object models and backgrounds, adaptation to facet or environment changes, good features for object representation and efficient use of a priori knowledge about object signatures [25].…”