Search

The morphological differentiation of bone marrow is fundamental for the diagnosis of leukemia. Currently, the counting and classification of the different types of bone marrow cells is done manually under the use of bright field microscopy. This is a time-consuming, subjective, tedious and error-prone process. Furthermore, repeated examinations of a slide may yield intra- and inter-observer variances. For that reason a computer assisted diagnosis system for bone marrow differentiation is pursued. In this work we focus (a) on a new method for the separation of nucleus and plasma parts and (b) on a knowledge-based hierarchical tree classifier for the differentiation of bone marrow cells in 16 different classes. Classification trees are easily interpretable and understandable and provide a classification together with an explanation. Using classification trees, expert knowledge (i.e. knowledge about similar classes and cell lines in the tree model of hematopoiesis) is integrated in the structure of the tree. The proposed segmentation method is evaluated with more than 10,000 manually segmented cells. For the evaluation of the proposed hierarchical classifier more than 140,000 automatically segmented bone marrow cells are used. Future automated solutions for the morphological analysis of bone marrow smears could potentially apply such an approach for the pre-classification of bone marrow cells and thereby shortening the examination time.

This paper investigates the effectiveness of an advanced classification system for accurate crop classification using very high resolution (VHR) satellite imagery. Specifically, a recently proposed genetic fuzzy rule-based classification system (GFRBCS) is employed, namely, the Hierarchical Rule-based Linguistic Classifier (HiRLiC). HiRLiC’s model comprises a small set of simple IF–THEN fuzzy rules, easily interpretable by humans. One of its most important attributes is that its learning algorithm requires minimum user interaction, since the most important learning parameters affecting the classification accuracy are determined by the learning algorithm automatically. HiRLiC is applied in a challenging crop classification task, using a SPOT5 satellite image over an intensively cultivated area in a lake-wetland ecosystem in northern Greece. A rich set of higher-order spectral and textural features is derived from the initial bands of the (pan-sharpened) image, resulting in an input space comprising 119 features. The experimental analysis proves that HiRLiC compares favorably to other interpretable classifiers of the literature, both in terms of structural complexity and classification accuracy. Its testing accuracy was very close to that obtained by complex state-of-the-art classification systems, such as the support vector machines (SVM) and random forest (RF) classifiers. Nevertheless, visual inspection of the derived classification maps shows that HiRLiC is characterized by higher generalization properties, providing more homogeneous classifications that the competitors. Moreover, the runtime requirements for producing the thematic map was orders of magnitude lower than the respective for the competitors.

We describe an architecture to provide online semantic labeling capabilities to field robots operating in urban environments. At the core of our system is the stacked hierarchical classifier developed by Munoz et al., which classifies regions in monocular color images using models derived from hand labeled training data. The classifier is trained to identify buildings, several kinds of hard surfaces, grass, trees, and sky. When taking this algorithm into the real world, practical concerns with difficult and varying lighting conditions require careful control of the imaging process. First, camera exposure is controlled by software, examining all of the image's pixels, to compensate for the poorly performing, simplistic algorithm used on the camera. Second, by merging multiple images taken with different exposure times, we are able to synthesize images with higher dynamic range than the ones produced by the sensor itself. The sensor 's limited dynamic range makes it difficult to, at the same time, properly expose areas in shadow along with high albedo surfaces that are directly illuminated by the sun. Texture is a key feature used by the classifier, and under /over exposed regions lacking texture are a leading cause of misclassifications. The results of the classifier are shared with higher-lev elements operating in the UGV in order to perform tasks such as building identification from a distance and finding traversable surfaces.

Web page classification is one of the essential techniques for Web mining. This paper proposes a binary hierarchical classifier for multi-class support vector machines for web page classification. This method applies truncated singular value decomposition on the training data that reduces its dimension and the noise data. After the truncated singular value decomposition on the training data, it uses the improved k-means algorithm design the binary hierarchical structure, the improved k-means algorithm makes the separability of one macro-class is the smallest, makes the separability of two macro-classes is the largest. The result of experiment performed on the training datasets shows that this algorithm can enhance precision of web page classification.

In CT angiography images, osseous structures occluding vessels pose dic ulties for physicians during diagnosis.Simple thresholding techniques for removing bones fail due to overlappingCT values of vessels “lled with contrastagent and osseous tissue, while manual delineation is slow and tedious. Thus, we propose to automaticallysegment bones using a trainable classi“er to label image patches as bone or background. The image featuresprovided to the classi“er are based on grey value statistics and gradients. In contrast to most existing methods,osseous tissue segmentation in our algorithm works without any prior knowledge of the body region depicted inthe image. This is achieved by using a probabilistic boosting tree, which is capable of automatically decomposingthe input space. The whole system works by partitioning the image using a watershed transform, classifyingimage regions as bone or background and re“ning the result by means of a graph-based procedure. Additionally,an intuitive way of manually re“ning the segmentation result is incorporated. The system was evaluated on 15CTA datasets acquired from various body regions, showing an average correct recognition of bone regions of 80%at a false positive rate of 0.025% of the background voxels.Keywords: CT angiography, CTA, bone removal, segmentation, hierarchical classi“er, boosting

We propose to use new SVM-type classifiers in a binary hierarchical tree classification structure to efficiently address the multi-class classification problem. A new hierarchical design method, WSV (weighted support vector) K-means Clustering, is presented; it automatically selects the classes to be separated at each node in the hierarchy. Our method is able to visualize and cluster high-dimensional support vector data; therefore, it improves upon prior hierarchical classifier designs. At each node in the hierarchy, we apply an SVRDM (support vector representation and discrimination machine) classifier, which offers generalization and good rejection of unseen false objects; rejection is not achieved with the standard SVM classifier. We provide the theoretical basis and insight into the choice of the Gaussian kernel to provide the SVRDM's rejection ability. New classification and rejection test results are presented on a real IR (infra-red) database.

We propose a binary hierarchical classifier to solve the multi-class classification problem with aspect variations in objects and with rejection of non-object false targets. The hierarchical architecture design is automated using our new k-means SVRM (support vector representation machine) clustering algorithm. At each node in the hierarchy, we use a new SVRDM (support vector representation and discrimination machine) classifier, which has good generalization and offers good rejection ability. We also provide a theoretical basis for our choice of kernel function (K), and our method of parameter selection (for σ and p ). Using this hierarchical SVRDM classifier with magnitude Fourier transform features, experimental results on both simulated and real infra-red (IR) databases are excellent.

Recently, these authors developed a heterogeneous, one-level image classifier (CH ) based on morphological descriptors from direct domain analysis (spatial differentiation), fractal analysis and "spectrum enhancement" (a kind of non-linear filtering). Classifier CH was applied to epi-fluorescence microscope images of cytoskeletal microtubules and was trained to recognize structural alterations of the cytoskeleton in various circumstances. The application dealt with images of rat hepatocytes (rh ). The scope of this paper is twofold: a ) to investigate different classifier architectures, which include the multiobjective optimization of some image analysis parameters by means of suitable algorithms; b ) to apply said classifiers to new sets of images obtained from mouse fibroblasts (mf ) and HepG2 (hg ) cells. Image sets from control and treated cell cultures are analyzed. Classifier CH is applied to mf microtubules. A new classifier entirely relying on "spectrum enhancement" (although on different descriptors) is developed and applied to rh and hg images. From the latter classifier, by bringing in descriptors from direct domain and fractal analysis, a hierarchical classifier is derived and applied to rh images. Results are compared. Classifier performance is expressed in terms of sensitivity, specificity and information contents of the first three principal components.© (2005) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

A fuzzy logic-based system to classify olfactive signals is presented. The odor samples are obtained from an electronic noise that contains conducting polymer sensors with partially overlapping sensitivities to odors. The sensor responses are represented by means of the coefficients of their Fast Fourier Transform (FFT). A feature reduction method is applied to reduce the feature space dimension. Then, an unsupervised Fuzzy Divisive Hierarchical Clustering (FDHC) method is used to establish the optimal number of clusters in the data set as well as the optimal cluster structure. The output of FDHC is a binary hierarchy of fuzzy classes that are used to build a supervised fuzzy hierarchical classifier. At each level of the fuzzy hierarchy a separating hyperplane of the two corresponding fuzzy training classes is determined. The hyperplane identifies two crisp decision regions, which will be refined at the next level of the hierarchy. In this way, we obtain a hierarchy of regions, which defines a crisp decision tree. Each region is, therefore, related to a specific expected output of the system. Recognition of an unknown odor is accomplished by computing the FFT of the corresponding signal and using the decision tree to establish the region the odor belongs to. Two small-scale applications of the method yielded 100% classification accuracy on out-of-sample data.© (1999) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Grouping images into (semantically) meaningful categories using low-level visual features is still a challenging and important problem in content-based image retrieval. Based on these groupings, effective indices can be built for an image database. In this paper, we cast the image classification problem in a Bayesian framework. Specifically, we consider city vs. landscape classification, and further, classification of landscape into sunset, forest, and mountain classes. We demonstrate how high-level concepts can be understood from specific low-level image features, under the constraint that the test images do belong to one of the delineated classes. We further demonstrate that a small codebook (the optimal size is selected using the MDL principle) extracted from a vector quantizer, can be used to estimate the class-conditional densities needed for the Bayesian methodology. Classification based on color histograms, color coherence vectors, edge direction histograms, and edge-direction coherence vectors as features shows promising results. On a database of 2,716 city and landscape images, our system achieved an accuracy of 95.3 percent for city vs. landscape classification. On a subset of 528 landscape images, our system achieves an accuracy of 94.9 percent for sunset vs. forest and mountain classification, and 93.6 percent for forest vs. mountain classification. Our final goal is to combine multiple 2- class classifiers into a single hierarchical classifier.

We report about a hierarchical design for extracting ship features and recognizing ships from SAR images, and which will eventually feed a multisensor data fusion system for airborne surveillance. The target is segmented from the image background using directional thresholding and region merging processes. Ship end-points are then identified through a ship centerline detection performed with a Hough transform. A ship length estimate is calculated assuming that the ship heading and/or the cross-range resolution are known. A high-level ship classification identifies whether the target belongs to Line (mainly combatant military ships) or Merchant ship categories. Category discrimination is based on the radar scatterers' distribution in 9 ship sections along the ship's range profile. A 3-layer neural network has been trained on simulated scatterers distributions and supervised by a rule- based expert system to perform this task. The NN 'smoothes out' the rules and the confidence levels on the category declaration. Line ship type (Frigate, Destroyer, Cruiser, Battleship, Aircraft Carrier) is then estimated using a Bayes classifier based on the ship length. Classifier performances using simulated images are presented.© (1998) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

A novel method of segmenting un-transcribed speech into sub-word acoustic units using neural tree networks (NTN) is presented. A NTN is a hierarchical classifier that combines the properties of decision trees and feed-forward neural networks. Segmentation of speech into sub-word units (of acoustic significance) is vital for large vocabulary, continuous speech recognition on un-transcribed speech database.© (1994) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

This paper introduces a robust method which can be used to recognize English characters. In our method, we first present an Invariant Matrix (IM) corresponding to a unique character image under the polar system. This kind of matrix has many properties and it is insensitive to image translation, scaling rotation and noise. On the basis of invariant matrix, a set of similar discriminant functions (SDF) of English characters are established. Then, a feature extraction and recognition method based on the SDF functions is proposed. Feature vectors extracted by our method are reliable and have the maximum similarity for the same class of English character samples. Finally, according to our recognition model, we design a hierarchical classifier to recognize English characters. Experimental results show that the SDF function based on IM is an efficient criterion for feature extraction of English characters and our recognition model can obtain the recognition accuracies of 100 percent for all English characters.© (1992) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

In practical pattern recognition problems, one-shot classifiers such as single feedforward neural networks trained by back-propagation may operate inefficiently in a complex pattern space and/or have unstable trained configurations. An alternative is a decision tree classifier. The authors report on the design, training, and accuracy of a hierarchical classifier implementing neural nets. Each nonterminal node is a separate feedforward neural network and is neither restricted to binary decisions nor limited to using only one feature to make those decisions. The features are pyramid data structures: identical texture parameters calculated across three different image resolutions about the training sites. In this application, results show a twenty percent relative increase in accuracy over the monolithic classifier.© (1991) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

A two-level feature extraction classifier using a geometrical-moment feature space is described for multi-class distortion-invariant pattern recognition. The first-level classifier provides object class and aspect estimates using multi-class Fisher projections and optimized two-class Fisher projections in a hierarchical classifier. Aspect estimates are provided from ratios of the computed moments. The second-level classifier provides the final class estimate, distortion parameter estimates and the confidence of the estimates. Extensive test results on a ship image database are presented.

Hierarchical classifiers are discussed and a new implementation in digital image processors with a pipeline architecture is presented. This implementation makes optimal use of the different layers of look up tables that generally exist in all machines with this architecture. Classification can be considered to occur at almost real time, as far as the unit structure of the tree associated to the classifier is concerned. As the hierarchical classifier makes optimal use of all the information available, fast decision rules can be applied in most of the nodes, thus reducing considerably the overall computational burden. An example is provided in the context of classification of multi-temporal remote sensed data. To apply the technique, the evaluation of the features must have been done previously.

A decision-net solution involving a novel hierarchical classifier and a set of multiple directed graphs, as well as a neural-net solution, are respectively presented for large-class problem and mixture problem treatments of imaging spectrometer data. The clustering method for hierarchical classifier design, when used with multiple directed graphs, yields an efficient decision net. New directed-graph rules for reducing local maxima as well as the number of perturbations required, and the new starting-node rules for extending the reachability and reducing the search time of the graphs, are noted to yield superior results, as indicated by an illustrative 500-class imaging spectrometer problem.

Numerical analysis techniques for the interpretation of high-resolution imaging-spectrometer data are described and demonstrated. The method proposed involves the use of (1) a hierarchical classifier with a tree structure generated automatically by a Fisher linear-discriminant-function algorithm and (2) a novel multiple-directed-graph scheme which reduces the local maxima and the number of perturbations required. Results for a 500-class test problem involving simulated imaging-spectrometer data are presented in tables and graphs; 100-percent-correct classification is achieved with an improvement factor of 5.