This paper primarily investigates the use of shape-based features by an Automatic Target Recognition (ATR) system to
classify various types of targets in Synthetic Aperture Radar (SAR) images. In specific, shapes of target outlines are
represented via Elliptical Fourier Descriptors (EFDs), which, in turn, are utilized as recognition features. According to
the proposed ATR approach, a segmentation stage first isolates the target region from shadow and ground clutter via a
sequence of fast thresholding and morphological operations. Next, a number of EFDs are computed that can sufficiently
describe the salient characteristics of the target outline. Finally, a classification stage based on an ensemble of Support
Vector Machines identifies the target with the appropriate class label. In order to experimentally illustrate the merit of
the proposed approach, SAR intensity images from the well-known Moving and Stationary Target Acquisition and
Recognition (MSTAR) dataset were used as 10-class and 3-class recognition problems. Furthermore, comparisons were
drawn in terms of classification performance and computational complexity to other successful methods discussed in the
literature, such as template matching methods. The obtained results portray that only a very limited amount of EFDs are
required to achieve recognition rates that are competitive to well-established approaches.
In this paper, we introduce a modification of the Fuzzy ARTMAP (FAM) neural network, namely, the Fuzzy ARTMAP with adaptively weighted distances (FAMawd) neural network. In FAMawd we substitute the regular L1-norm with a weighted L1-norm to measure the distances between categories and input patterns. The distance-related weights are a function of a category's shape and allow for bias in the direction of a category's expansion during learning. Moreover, the modification to the distance measurement is proposed in order to study the capability of FAMawd in achieving more compact knowledge representation than FAM, while simultaneously maintaining good classification performance. For a special parameter setting FAMawd simplifies to the original FAM, thus, making FAMawd a generalization of the FAM architecture. We also present an experimental comparison between FAMawd and FAM on two benchmark classification problems in terms of generalization performance and utilization of categories. Our obtained results illustrate FAMawd's potential to exhibit low memory utilization, while maintaining classification performance comparable to FAM.
Ellipsoid ARTMAP (EAM) is an adaptive-resonance-theory neural network architecture that is capable of successfully performing classification tasks using incremental learning. EAM achieves its task by summarizing labeled input data via hyper-ellipsoidal structures (categories). A major property of EAM, when using off-line fast learning, is that it perfectly learns its training set after training has completed. Depending on the classification problems at hand, this fact implies that off-line EAM training may potentially suffer from over-fitting. For such problems we present an enhancement to the basic Ellipsoid ARTMAP architecture, namely Boosted Ellipsoid ARTMAP (bEAM), that is designed to simultaneously improve the generalization properties and reduce the number of created categories for EAM's off-line fast learning. This is being accomplished by forcing EAM to be tolerant about occasional misclassification errors during fast learning. An additional advantage provided by bEAM's desing is the capability of learning inconsistent cases, that is, learning identical patterns with contradicting class labels. After we present the theory behind bEAM's enhancements, we provide some preliminary experimental results, which compare the new variant to the original EAM network, Probabilistic EAM and three different variants of the Restricted Coulomb Energy neural network on the square-in-a-square classification problem.
In the recent past category regions have been introduced as new geometrical concepts and provide a visualization tool that facilitates significant insight into the nature of the competition among categories during both the training and performance phase of Fuzzy ART (FA) and Fuzzy ARTMAP (FAM). These regions are defined as the geometric interpretation of the Vigilance Test and the competition of each category with an uncommitted F2-layer node for a specific input pattern (Commitment Test). In this paper we show how the notion of category regions can be naturally extended to Ellipsoid ART (EA) and Ellipsoid ARTMAP (EAM) and focus on the regions' theoretical properties, when considering the Choice-by-Difference category choice function. Based on these properties we state three theoretical results applicable to both EA and EAM. Specifically, if r and a denote the vigilance and the choice parameter respectively, we show that in certain areas of the (a,r) plane the result of EA/EAM training is independent of the specific value of either r or w (parameter of the activation function value for an uncommitted F2-layer node). Finally, we provide a refined upper bound on the size of categories created in EA/EAM during training. All the results are immediately applicable to FA/FAM as well.
In this paper we are examining the issue of overtraining in Fuzzy ARTMAP. Over-training in Fuzzy ARTMAP manifests itself in two different ways: (a) it degrades the generalization performance of Fuzzy ARTMAP as training progresses, and (b) it creates unnecessarily large Fuzzy ARTMAP neural network architectures. In this work we are demonstrating that overtraining happens in Fuzzy ARTMAP and we propose an old remedy for its cure: cross-validation. In our experiments we compare the performance of Fuzzy ARTMAP that is trained (i) until the completion of training, (ii) for one epoch, and (iii) until its performance on a validation set is maximized. The experiments were performed on artificial and real databases. The conclusion derived from these experiments is that cross-validation is a useful procedure in Fuzzy ARTMAP, because it produces smaller Fuzzy ARTMAP architectures with improved generalization performance. The trade-off is that cross-validation introduces additional computational complexity in the training phase of Fuzzy ARTMAP.
In this paper we introduce new useful, geometric concepts regarding categories in Fuzzy ART and Fuzzy ARTMAP, which shed more light into the process of category competition eligibility upon the presentation of input patterns. First, we reformulate the competition of committed nodes with uncommitted nodes in an F2 layer as a commitment test very similar to the vigilance test. Next, we introduce a category's match and choice regions, which are the geometric interpretation of the vigilance and commitment test respectively. After examining properties of these regions we reach three results applicable to both Fuzzy ART and Fuzzy ARTMAP. More specifically, we show that only one out of these two tests is required; which test needs to be performed depends on the values of the vigilance parameter (rho) and the choice parameter (alpha) . Also, we show that for a specific relation of (rho) and (alpha) , the vigilance (rho) does not influence the training or performance phase of Fuzzy ART and Fuzzy ARTMAP. Finally, we refine a previously published upper bound on the size of the categories created during training in Fuzzy ART and Fuzzy ARTMAP.
We introduce Ellipsoid-ART, EA and Ellipsoid-ARTMAP, EAM as a generalization of Hyper-sphere ART and Hypersphere-ARTMAP respectively. Our novel archetectures are based on ideas rooted on Fuzzy-ART, FA and Fuzzy-ARTMAP, FAM. While FA/FAM summarize input data using hyper-rectangles, EA/EAM utilize hyper-ellipsoids for the same purpose. Due to their learning rules, EA and EAM share virtually all properties and characteristics of their FA/FAM counterparts. Preliminary experimentation implies that EA and EAM are to be viewed as good alternatives to FA and FAM for data clustering and classification tasks. Extensive pseudo-code is provided in the appendices for computationally efficient implementations of EA/EAM training and performance phases.
In this paper we present a modification of the test phase of ARTMAP-based neural networks that improves the classification performance of the networks when the patterns that are used for classification are extracted from noisy signals. The signals that are considered in this work are textured images, which are a case of 2D signals. Two neural networks from the ARTMAP family are examined, namely the Fuzzy ARTMAP (FAM) neural network and the Hypersphere ARTMAP (HAM) neural network. We compare the original FAM and HAM architectures with the modified ones, which we name FAM-m and HAM-m respectively. We also compare the classification performance of the modified networks, and of the original networks when they are trained with patterns extracted from noisy textures. Finally, we illustrate how combination of features can improve the classification performance for both the noiseless and noisy textures.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.