Soft Classification Research Articles

Maximisation of mutual information for gait-based soft biometric classification using Gabor features

Besides identity, soft biometric characteristics, such as gender and age can also be derived from gait patterns. With Gabor enhancement, supervised learning and temporal modelling, the authors present a robust framework to achieve state-of-the-art classification accuracy for both gender and age. Gabor filter and maximisation of mutual information are used to extract low-dimensional features, whereas Bayes rules based on hidden Markov models (HMMs) are adopted for soft biometric classification. The multi-view soft biometric classification problem is defined as two different cases, saying, one-to-one view and many-to-one view, according to the number of available gallery views. In case more than one gallery view is available, the multi-view soft biometric classification problem is hierarchically solved with a view-related population HMM, in which the estimated view angle is treated as the intermediate result in the first stage. Performance has been evaluated on benchmark databases, which verify the advantages of the proposed algorithm.

Prostate segmentation by sparse representation based classification.

Accurate segmentation of prostate in CT images is important in image-guided radiotherapy. However, it is difficult to localize the prostate in CT images due to low image contrast, unpredicted motion and large appearance variations across different treatment days. To address these issues, we propose a sparse representation based classification method to accurately segment the prostate. The main contributions of this paper are: (1) A discriminant dictionary learning technique is proposed to overcome the limitation of the traditional Sparse Representation based classifier (SRC). (2) Context features are incorporated into SRC to refine the prostate boundary in an iterative scheme. (3) A residue-based linear regression model is trained to increase the classification performance of SRC and extend it from hard classification to soft classification. To segment the prostate, the new treatment image is first rigidly aligned to the planning image space based on the pelvic bones. Then two sets of location-adaptive SRCs along two coordinate directions are applied on the aligned treatment image to produce a probability map, based on which all previously segmented images of the same patient are rigidly aligned onto the new treatment image and majority voting strategy is further adopted to finally segment the prostate in the new treatment image. The proposed method has been evaluated on a CT dataset consisting of 15 patients and 230 CT images. Promising results have been achieved.

Super-resolution land-cover mapping based on the selective endmember spectral mixture model in hyperspectral imagery

Hyperspectral imagery contains a large number of mixed pixels, which limits its utility. Super-resolution mapping is a potential solution to this problem, designed to use the proportion of land covers to obtain a sharpened thematic map with higher resolution. Endmember is a fundamental variable in the process, which is a critical issue for decomposing the mixed pixels and sharpening the subpixel level images. In most cases, the forms of the endmember combination in diverse pixels are very distinct. However, traditional soft classification methods neglect this point and model endmembers as fixed composition entities. Due to the reliance on this flawed spectral mixture model, the super-resolution mapping is unable to represent detail in the following result image precisely and effectively. In this work, therefore, endmember variability is considered, focusing on identifying the most suitable form of the endmember combination. This issue is addressed by applying a new selective endmember spectral mixture (SESM) model, which allows the endmember number and type to vary at a per pixel level, and then super-resolution mapping can be subsequently performed according to the produced spectral abundances. Two different types of hyperspectral data are used in our experiments. First, the SESM model is tested individually for validation of its applicability. Then the complete algorithm integrating SESM and super-resolution mapping based on a back-propagation neural network is evaluated. It showed that a more accurate endmember combination in the parent pixel results in a finer representation image. The experimental results prove that the proposed algorithm can effectively improve the accuracy of the super-resolution mapping results compared to the traditional method.

Improvement of classification using robust soft classification rules for near-infrared reflectance spectral data

The aim of this work was to propose a quick and cost-effective procedure, which could help to identify the types of fat (rapeseed, a mixture of rapeseed and soybean, and lard oils) added to feed used for raising pigs. For this purpose, liver samples were examined and their near-infrared reflectance spectra served as data for the construction of classic and robust soft independent modeling of class analogy (SIMCA) models. The results showed that the near-infrared reflectance spectra contained information sufficient to build good classification models that enabled three types of fat additions to be distinguished. The best classification results were obtained from robust SIMCA, indicating its superior performance in terms of high sensitivity and specificity in comparison with classic SIMCA. Specifically, robust models had sensitivities of 100% and specificities of 96.05%, 97.73% and 100%, for rapeseed, mixture of rapeseed and soybean, and lard enriched feed, respectively.

Possibilities and limitations of artificial neural networks for subpixel mapping of land cover

Although developments in remote sensing have greatly improved land cover mapping, the mixed pixel problem has not yet been fully addressed. Soft classification techniques have been introduced to address the problem, but they do not show the spatial location of the class proportions in a pixel. Subpixel mapping has been introduced to address the drawbacks of soft classifications. In this work, the feedforward backpropagating neural network (FFBPNN) was used for subpixel mapping. A set of class proportion images, which are to be treated as soft classification results, were created from a high spatial resolution (25 m) land cover dataset. For this purpose, the land cover dataset was aggregated both thematically (into two, four or eight land cover classes) and spatially (into proportion images with pixel sizes of 75, 150 and 300 m). This resulted in nine different combinations that were considered here as study cases. Several FFBPNNs were trained using these proportion images and the original land cover dataset (which was used as a target). Subsequently, the best networks were used to reconstruct high spatial resolution land cover maps of two heterogeneous areas in the south of The Netherlands. The overall accuracies obtained revealed that the networks were influenced by the spatial frequency, shape and size of the different land cover types. Moreover, it was revealed that most of the errors were on the class boundaries where highly mixed pixels are to be expected. The accuracies spanned a wide range of values depending on the complexity of the cases. Although it was not possible to exhaustively explore all network architectures, the results demonstrate the potential of the FFBPNN for subpixel mapping.

Super-resolution mapping using Hopfield Neural Network with panchromatic imagery

Land-cover proportions of mixed pixels can be predicted using soft classification. From the land-cover proportions, a hard land-cover map can be predicted at sub-pixel spatial resolution using super-resolution mapping techniques. It has been demonstrated that the Hopfield Neural Network (HNN) provides a suitable method for super-resolution mapping. To increase the detail and accuracy of the sub-pixel land-cover map, supplementary information at an intermediate spatial resolution can be used. In this research, panchromatic (PAN) imagery was used as an additional source of information for super-resolution mapping. Information from the PAN image was captured by a new PAN reflectance constraint in the energy function of the HNN. The value of the new PAN reflectance constraint was defined based on forward and inverse models with local end-member spectra and local convolution weighting factors. Two sets of simulated and degraded data were used to test the new technique. The results indicate that PAN imagery can be used as a source of supplementary information to increase the detail and accuracy of sub-pixel land-cover maps produced by super-resolution mapping from land-cover proportion images.

Dealing with vagueness in complex forest landscapes: A soft classification approach through a niche-based distribution model

The increasing interest in biodiversity conservation has led to the development of new approaches to facilitate ecologically based conservation policies and management plans. In this context, the development of effective methods for the classification of forest types constitutes a crucial issue as forests represent the most widespread vegetation structure and play a key role in ecosystem functioning. In this study a maximum entropy approach (Maxent) to forest type classification in a complex Mediterranean area, has been investigated. Maxent, a niche-based model of species/habitat distribution, allowed researchers to estimate the potential distribution of four forest types: Holm oak, Mixed oak, Mixed broadleaved and Riparian forests. The Maxent model's internal tests have proved a powerful tool for estimating the model's accuracy and analyzing the effects of the most important variables in the produced models. Moreover the comparison with a spectral response-based fuzzy classification, showed a higher accuracy in the Maxent outputs, demonstrating how the use of environmental variables, combined with spectral information in the classification of natural or semi-natural land cover classes, improves map accuracies. The modeling approach followed by this study, taking into account the uncertainty proper of the natural ecosystems and the use of environmental variables in land cover classification, can represent a useful approach to making more efficient and effective field inventories and to developing effective conservation policies.

Characterizing spatial patterns of invasive species using sub-pixel classifications

Invasive species disrupt landscape patterns and compromise the functionality of ecosystem processes. Non-native saltcedar poses significant threats to native vegetation and groundwater resources in the southwestern U.S. and Mexico, and quantifying spatial and temporal distribution patterns is essential for monitoring its spread. Considerable research focuses on determining the accuracy of various remote sensing techniques for distinguishing saltcedar from native woody riparian vegetation through sub-pixel, or soft classifications. However, there is a lack of research quantifying spatial distribution patterns from these classifications, mainly because landscape metrics, which are commonly used to statistically assess these patterns, require bounded classes and cannot be applied directly to soft classifications. This study tests a new method for discretizing sub-pixel data to generate landscape metrics using a continuum of fractional cover thresholds. The developed approach transforms sub-pixel classifications into discrete maps compliant with metric terms and computes and interprets metric results in the context of the region to explain patterns in the extent, distribution, and connectivity of saltcedar in the Rio Grande basin. Results indicate that landscape metrics are sensitive to sub-pixel values and can vary greatly with fractional cover. Therefore spectral unmixing should be performed prior to metric calculations. Analysis of metric trends provides evidence that saltcedar has expanded away from the immediate riparian zones and is displacing native vegetation. This information, coupled with control management strategies, can be used to target remediation activities along the Rio Grande.

Efficient collection of training data for sub-pixel land cover classification using neural networks

Artificial neural networks (ANNs) are a popular class of techniques for performing soft classifications of satellite images. They have successfully been applied for estimating crop areas through sub-pixel classification of medium to low resolution images. Before a network can be used for classification and estimation, however, it has to be trained. The collection of the reference area fractions needed to train an ANN is often both time-consuming and expensive. This study focuses on strategies for decreasing the efforts needed to collect the necessary reference data, without compromising the accuracy of the resulting area estimates. Two aspects were studied: the spatial sampling scheme (i) and the possibility for reusing trained networks in multiple consecutive seasons (ii). Belgium was chosen as the study area because of the vast amount of reference data available. Time series of monthly NDVI composites for both SPOT-VGT and MODIS were used as the network inputs. The results showed that accurate regional crop area estimation (R2>80%) is possible using only 1% of the entire area for network training, provided that the training samples used are representative for the land use variability present in the study area. Limiting the training samples to a specific subset of the population, either geographically or thematically, significantly decreased the accuracy of the estimates. The results also indicate that the use of ANNs trained with data from one season to estimate area fractions in another season is not to be recommended. The interannual variability observed in the endmembers’ spectral signatures underlines the importance of using up-to-date training samples. It can thus be concluded that the representativeness of the training samples, both regarding the spatial and the temporal aspects, is an important issue in crop area estimation using ANNs that should not easily be ignored.

The impact of image and class structure upon sub-pixel mapping accuracy using the pixel-swapping algorithm

A series of tests assess the impact of image and class structure within input imagery upon the accuracy of outputs from the pixel-swapping algorithm (PSA) to better assess the viability of this sub-pixel mapping algorithm to spatially locate class proportions and improve soft classification accuracies. Simulated maps with known spatial autocorrelation, class proportion and mixed pixel percentages are sub-pixel mapped at varying zoom factors using the PSA and accuracies measured. Block majority filtering (BMF) of input images is an experimental accuracy control. Results quantify PSA accuracies compared to the BMF based on the interrelationship between spatial autocorrelation, class proportions, zoom factor and mixed pixel quantities. Findings, while validating the known positive correlation between spatial autocorrelation and PSA accuracy, demonstrate class proportion difference and mixed pixel quantities to be positively and negatively related to PSA accuracy, respectively. Class proportion affects PSA accuracy to a greater degree on maps exhibiting low spatial autocorrelation levels versus maps with high spatial autocorrelation levels. Traditional accuracy measures, such as overall accuracy, are shown to overestimate the true performance of the PSA. The difference in overall accuracy of the PSA between all map pixels and within mixed map pixels alone can be over 20%. Overall, in general, the PSA accurately sub-pixel mapped input images with an approximate minimum level of spatial autocorrelation at a Moran's I of 0.63, a maximum of 30% of mixed pixels relative to entire image pixels, a difference in class proportions greater than 40% and at zoom factors below 10.

Spectral Unmixing for the Classification of Hyperspectral Images at a Finer Spatial Resolution

The problem of classification of hyperspectral images containing mixed pixels is addressed. Hyperspectral imaging is a continuously growing area of remote sensing applications. The wide spectral range of such imagery, providing a very high spectral resolution, allows to detect and classify surfaces and chemical elements of the observed image. The main problem of hyperspectral data is the (relatively) low spatial resolution, which can vary from a few to tens of meters. Many factors make the spatial resolution one of the most expensive and hardest to improve in imaging systems. For classification, the major problem caused by low spatial resolution are the mixed pixels, i.e., parts of the image where more than one land cover map lie in the same pixel. In this paper, we propose a method to address the problem of mixed pixels and to obtain a finer spatial resolution of the land cover classification maps. The method exploits the advantages of both soft classification techniques and spectral unmixing algorithms, in order to determine the fractional abundances of the classes at a sub-pixel scale. Spatial regularization by simulated annealing is finally performed to spatially locate the obtained classes. Experiments carried out on synthetic real data sets show excellent results both from a qualitative and quantitative point of view.

Hard or Soft Classification? Large-Margin Unified Machines

Margin-based classifiers have been popular in both machine learning and statistics for classification problems. Among numerous classifiers, some are hard classifiers while some are soft ones. Soft classifiers explicitly estimate the class conditional probabilities and then perform classification based on estimated probabilities. In contrast, hard classifiers directly target the classification decision boundary without producing the probability estimation. These two types of classifiers are based on different philosophies and each has its own merits. In this article, we propose a novel family of large-margin classifiers, namely large-margin unified machines (LUMs), which covers a broad range of margin-based classifiers including both hard and soft ones. By offering a natural bridge from soft to hard classification, the LUM provides a unified algorithm to fit various classifiers and hence a convenient platform to compare hard and soft classification. Both theoretical consistency and numerical performance of LUMs are explored. Our numerical study sheds some light on the choice between hard and soft classifiers in various classification problems.

Simultaneous latent fingerprint recognition

Abstract Simultaneous latent fingerprints are clusters of friction ridge impressions deposited concurrently in a crime scene. The analysis of these impressions is a complex task to infer individualization, exclusion or categorize as inconclusive. The problem is further compounded when distinctive features in each latent fingerprint in the cluster are of varying quality or none of the fingerprint has the requisite number of features to reliably arrive at a conclusion. Recently, SWGFAST (Scientific Working Group on Friction Ridge Analysis, Study and Technology) proposed a draft standard for simultaneous impression examination. The approach is manual and requires known reference ten-print for comparing with an unknown simultaneous latent fingerprint. This paper proposes a semi-automatic approach to process and analyze simultaneous latent fingerprints. The proposed algorithm demonstrates that comparisons can be made from a database of ten-prints for a more comprehensive search instead of the time consuming manual approach used by latent fingerprint examiners. The algorithm was implemented using several soft computing and classification approaches and the performance was compared using the simultaneous latent fingerprint database. The results show that 2 ν-SVM with RBF kernel gave the best results both in terms of time and accuracy.

Using a sub-pixel mapping model to improve the accuracy of landscape pattern indices

The assessment of landscape spatial patterns is a key issue in landscape management. Landscape pattern indices (LPIs) are tools appropriate for analyzing landscape spatial patterns. LPIs are often derived from raster land cover maps that are extracted from remotely sensed data through hard classification. However, pixel-based hard classification methods suffer from the mixed pixel problem (in which pixels contain more than one land cover class), making for inaccurate classification maps and LPIs. In addition, LPIs generated by hard classification methods are characterized by grain sizes (the sampling unit sizes) that limit the derived landscape pattern to a certain scale. Sub-pixel mapping (SPM) models can enable fine-scale estimation of the spatial patterns of land cover classes without requiring additional data; hence, this is an appropriate downscaling method for land cover mapping. The fraction images generated by soft classification estimate the area proportion of each land cover class within each pixel, and using these images as input enables SPM models to alleviate the mixed pixel problem. At the same time, by transforming fraction images into a finer-scaled hard classification map, SPM models can minimize the influence of grain size on LPIs calculation. In this research, simulated landscape thematic patterns that can provide different landscape spatial patterns, eight commonly used LPIs and a SPM model that maximizes the spatial dependence between neighbouring sub-pixels were applied to assess the efficiency of deriving LPIs from sub-pixel model maps. Results showed that the SPM model can more precisely characterize landscape patterns than hard classification methods can. Landscape fragmentation, class abundance, the uncertainty in SPM, and the spatial resolution of the remotely sensed data influenced LPIs derived from sub-pixel maps. The largest patch index, landscape division, and patch cohesion derived from remotely sensed data with different spatial resolutions through the SPM model were suitable for inter-comparison, whereas the patch density, mean patch area, edge density, landscape shape index, and area-weighted mean shape index derived from the sub-pixel maps were sensitive to the spatial resolution of the remotely sensed data.

A Phenotypic Assessment Tool for Craniofacial Microsomia

Craniofacial microsomia is one of the most common conditions treated by craniofacial teams. However, research regarding the cause of this condition or the surgical outcomes of treatment is scant. This is attributable to the lack of diagnostic criteria and the wide phenotypic spectrum. Standardized description of the craniofacial malformations associated with craniofacial microsomia is a necessary first step for multicenter, interdisciplinary research into this complex condition. The authors used the previously published pictorial Orbit, Mandible, Ear, Nerve, and Soft tissue-Plus classification scheme to assign a phenotypic severity score to patients with craniofacial microsomia treated at the Craniofacial Center at Seattle Children's Hospital. The authors modified the tool based on feedback from multidisciplinary focus groups. The authors also developed a standardized photographic protocol to facilitate assessment of patients using two-dimensional images. Feedback from focus groups was synthesized to create a phenotypic assessment tool for craniofacial microsomia based on the pictorial Orbit, Mandible, Ear, Nerve, and Soft tissue-Plus classification system. This tool allows for more comprehensive description of the phenotype of craniofacial microsomia and is found to be effective for clinical use within a multidisciplinary craniofacial team. In addition, the photographic protocol for patients with craniofacial microsomia allows for classification from a two-dimensional photographic database, thereby facilitating research using archived photographs. The phenotypic assessment tool for craniofacial microsomia protocol provides a simple and standardized method for practitioners and researchers to classify patients with craniofacial microsomia. We anticipate that this tool can be used in multicenter investigational studies to evaluate the cause of this condition, its natural history, and comparative effectiveness research.

Hard or Soft Classification? Large-Margin Unified Machines

Hard or Soft Classification? Large-Margin Unified Machines

A Soft Classification Model for Vendor Selection

A Soft Classification Model for Vendor Selection

Soft classification of mixed seabed objects based on fuzzy clustering analysis using airborne LIDAR bathymetry data

Coastal seabed mapping is essential for a variety of nearshore management related activities including sustainable resource management, ecological protection, and environmental change detection in coastal sites...In this study, a soft seabed classification (SSC) algorithm is developed using unsupervised classification with fuzzy clustering to produce classification products accounting for a sub-footprint habitat mixture. With this approach, each shot is assigned to multiple seabed classes with a percentage cover measuring the extent to which each seabed class is present in the ALB footprint. This has the added benefit of generating smooth spatial ecological transitions of the seabed instead of sharp boundaries between classes or clusters. Furthermore, due to the multivariate nature of the SSC output (i.e., percentage cover for each seabed class for a given shot), a recently developed selforganizing map neural network-based approach to geo-visualization of seabed classification results was used to visualize seabed habitat diversity. An ALB dataset of an area approximately 20000 m2 collected from Quebec, Canada was used. Cross-validation of the SSC approach yields percentage cover accuracy of approximately 71.7% with 16 seabed classes for a real ALB dataset, while dominant seabed class prediction based on hardening of percentage cover predictions yielded 66% accuracy for 4 seabed classes.

Comparing the performance of fuzzy and crisp classifiers on remotely sensed images: a case of snow classification

This study deals with the evaluation of accuracy benefits offered by a fuzzy classifier as compared to hard classifiers using satellite imagery for thematic mapping applications. When a crisp classifier approach is adopted to classify moderate resolution data, the presence of mixed coverage pixels implies that the final product will have errors, either of omission or commission, which are not avoidable and are solely due to the spatial resolution of the data. Theoretically, a soft classifier is not affected by such errors, and in principle can produce a classification that is more accurate than any hard classifier. In this study we use the Pareto boundary of optimal solutions as a quantitative method to compare the performance of a fuzzy statistical classifier to the one of two hard classifiers, and to determine the highest accuracy which could be achieved by hard classifiers. As an application, the method is applied to a case of snow mapping from Moderate-Resolution Imaging Spectroradiometer (MODIS) data on two alpine sites, validated with contemporaneous fine-resolution Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) data. The results for this case study showed that the soft classifier not only outperformed the two crisp classifiers, but also yielded higher accuracy than the maximum theoretical accuracy of any crisp classifier on the study areas. While providing a general assessment framework for the performance of soft classifiers, the results obtained by this inter-comparison exercise showed that soft classifiers can be an effective solution to overcome errors which are intrinsic in the classification of coarse and moderate resolution data.

Spatio-temporal pattern of urban land cover evolvement with urban renewal and expansion in Shanghai based on mixed-pixel classification for remote sensing imagery

Research into pixel unmixing in remote sensing imagery led to the development of soft classification methods. In this article, we propose a possibilistic c repulsive medoids (PCRMdd) clustering algorithm which attempts to find c repulsive medoids as a minimal solution of a particular objective function. The PCRMdd algorithm is applied to predict the proportion of each land use class within a single pixel, and generate a set of endmember fraction images. The clustering results obtained on multi-temporal Landsat Thematic Mapper (TM)/Enhanced Thematic Mapper plus (ETM+) images of Shanghai city in China reveal the spatio-temporal pattern of Shanghai land use evolvement and urban land spatial sprawl in course of urbanization from 1989 to 2002. The spatial pattern of land use transformation with urban renewal and expansion indicates the urban land use structure is gradually optimized during vigorous urban renewal and large-scale development of Pudong area, which will have an active influence on improving urban space landscape and enhancing the quality of the ecological environment. In addition, accuracy analysis demonstrates that PCRMdd represents a robust and effective tool for mixed-pixel classification on remote sensing imagery to obtain reliable soft classification results and endmember spectral information in a noisy environment.