Typical advantages and limitations of prior knowledge-based (deductive, top-down) expert systems are well known in literature: they typically score “high” in efficiency and interpretability, but they tend to score “low” in transferability/robustness to changes in input data. To benefit from these advantages while overcoming their typical shortcomings, an original expert system, based on a priori purely spectral-domain knowledge, is proposed for per-pixel (spatial context-insensitive) automatic near real-time detection of thermal anomalies in geostationary GOES-16 ABI multi-spectral (MS) imagery. Unable to learn-from-data, the proposed static decision-tree for MS signature recognition (classification) requires neither training data nor human-machine interaction to run. Its degrees of novelty pertain to the Marr levels of system understanding known as information/knowledge representation, system design (architecture) and implementation. Input with day and night ABI imagery acquired every 15 minutes, the proposed expert system detected 680 pixels with thermal anomalies in ABI images of the North and South Americas acquired from 30/01/2018 (15:00 UTC) to 31/01/2018 (01:30 UTC).

Within the 2015 IEEE GRSS Data Fusion Contest, an extremely high-resolution 3D LiDAR point cloud of a harbor test site must be “fused” with a 2D multi-spectral aerial image, featuring no radiometric calibration metadata file, of the same surface area. In this scenario we propose an innovative geospatial 2D and 3D object-based classification system, capable of counting instances of two populations of ISO-containers, whose standard dimensions are known a priori based on the ISO 668 - Series 1 freight containers documentation, detected in the 2D and 3D datasets at hand. The degree of novelty of the proposed classification system is twofold. First, it combines inductive (bottom-up, data-driven) and deductive (top-down, prior knowledge-based) inference mechanisms, where the latter initializes the former in a hybrid inference framework. Second, it is provided with feedback loops, which increase its robustness to changes in input data and augment its degree of automation. The geospatial outcome consists of tangible vector objects, which allow estimation of statistics per container together with a detailed reconstruction of the 3D scene in a geographic information system.

To deliver sample estimates provided with the necessary probability foundation to permit generalization from the sample data subset to the whole target population being sampled, probability sampling strategies are required to satisfy three necessary not sufficient conditions: 1) All inclusion probabilities be greater than zero in the target population to be sampled. If some sampling units have an inclusion probability of zero, then a map accuracy assessment does not represent the entire target region depicted in the map to be assessed. 2) The inclusion probabilities must be: a) knowable for nonsampled units and b) known for those units selected in the sample: since the inclusion probability determines the weight attached to each sampling unit in the accuracy estimation formulas, if the inclusion probabilities are unknown, so are the estimation weights. This original work presents a novel (to the best of these authors' knowledge, the first) probability sampling protocol for quality assessment and comparison of thematic maps generated from spaceborne/airborne very high resolution images, where: 1) an original Categorical Variable Pair Similarity Index (proposed in two different formulations) is estimated as a fuzzy degree of match between a reference and a test semantic vocabulary, which may not coincide, and 2) both symbolic pixel-based thematic quality indicators (TQIs) and sub-symbolic object-based spatial quality indicators (SQIs) are estimated with a degree of uncertainty in measurement in compliance with the well-known Quality Assurance Framework for Earth Observation (QA4EO) guidelines. Like a decision-tree, any protocol (guidelines for best practice) comprises a set of rules, equivalent to structural knowledge, and an order of presentation of the rule set, known as procedural knowledge. The combination of these two levels of knowledge makes an original protocol worth more than the sum of its parts. The several degrees of novelty of the proposed probability sampling protocol are highlighted in this paper, at the levels of understanding of both structural and procedural knowledge, in comparison with related multi-disciplinary works selected from the existing literature. In the experimental session, the proposed protocol is tested for accuracy validation of preliminary classification maps automatically generated by the Satellite Image Automatic Mapper (SIAM™) software product from two WorldView-2 images and one QuickBird-2 image provided by DigitalGlobe for testing purposes. In these experiments, collected TQIs and SQIs are statistically valid, statistically significant, consistent across maps, and in agreement with theoretical expectations, visual (qualitative) evidence and quantitative quality indexes of operativeness (OQIs) claimed for SIAM™ by related papers. As a subsidiary conclusion, the statistically consistent and statistically significant accuracy validation of the SIAM™ pre-classification maps proposed in this contribution, together with OQIs claimed for SIAM™ by related works, make the operational (automatic, accurate, near real-time, robust, scalable) SIAM™ software product eligible for opening up new inter-disciplinary research and market opportunities in accordance with the visionary goal of the Global Earth Observation System of Systems initiative and the QA4EO international guidelines.

Proposed in recent literature, a novel two-stage stratified hierarchical hybrid remote-sensing image understanding system (RS-IUS) architecture comprises the following: 1) a first-stage pixel-based application-independent top-down (physical-model-driven and prior-knowledge-based) preliminary classifier and 2) a second-stage battery of stratified hierarchical context-sensitive application-dependent modules for class-specific feature extraction and classification. The first-stage preliminary classifier is implemented as an operational automatic near-real-time per-pixel multisource multiresolution application-independent spectral-rule-based decision-tree classifier (SRC). To the best of the author's knowledge, SRC provides the first operational example of an automatic multisensor multiresolution Earth-observation (EO) system of systems envisaged under ongoing international research programs such as the Global Earth Observation System of Systems (GEOSS) and the Global Monitoring for the Environment and Security (GMES). For the sake of simplicity, the original SRC formulation adopts crisp (hard) membership functions unsuitable for dealing with component cover classes of mixed pixels (class mixture). In this paper, the crisp (hierarchical) SRC first stage of a two-stage hybrid RS-IUS is replaced by a fuzzy (horizontal) SRC. In operational terms, a relative comparison of the fuzzy SRC against its crisp counterpart reveals that the former features the following: 1) the same degree of automation which cannot be surpassed, i.e., they are both “fully automatic”; 2) a superior map information/knowledge representation where component cover classes of mixed pixels are modeled; 3) the same robustness to changes in the input multispectral imagery acquired across time, space, and sensors; 4) a superior maintainability/scalability/reusability guaranteed by an internal horizontal (flat) modular structure independent of hierarchy; and 5) a computation time increased by 30% in a single-process single-thread implementation. This computation overload would reduce to zero in a single-process multithread implementation. In line with theory, the conclusion of this work is that the operational qualities of the fuzzy and crisp SRCs differ, but both SRCs are suitable for the development of operational automatic near-real-time multisensor satellite-based measurement systems such as those conceived as a visionary goal by the ongoing GEOSS and GMES research initiatives.

To date, the automatic or semiautomatic transformation of huge amounts of multisource multiresolution spaceborne imagery into information still remains far below reasonable expectations. The original contribution of this paper to existing knowledge on the development of operational automatic remote sensing image understanding systems (RS-IUSs) is fourfold. First, existing RS-IUS architectures are critically revised. In this review section, the two-stage stratified hierarchical RS-IUS model, originally proposed by Shackelford and Davis, is identified as a subclass of the parent class of multiagent hybrid systems for RS image understanding, which is potentially superior to the two-stage segment-based RS-IUS architecture that is currently considered the state-of-the-art in commercial RS image-processing software toolboxes. Second, this paper highlights the degree of novelty of an operational automatic near-real-time well-posed model-driven application-independent per-pixel Landsat-like spectral-rule-based decision-tree classifier (LSRC) recently presented in RS literature. Third, five original downscaled implementations of the LSRC system are proposed to be input with a multispectral image whose spectral resolution overlaps with, but is inferior to, Landsat's. These five downscaled LSRC implementations are identified as the Satellite Pour l'Observation de la Terre-like SRC, the Advanced Very High Resolution Radiometer-like SRC, the Advanced Along-Track Scanning Radiometer-like SRC, the IKONOS-like SRC, and the Disaster Monitoring Constellation-like SRC, respectively. LSRC, together with its five downscaled implementations, called the integrated SRC system of systems, is eligible for use as the automatic pixel-based preliminary classification first stage of a two-stage stratified hierarchical RS-IUS instantiation. Fourth, to sustain the feasibility of the new downscaled LSRC implementations, a novel vegetation spectral index is introduced and discussed. In Part II of this paper, experimental results are presented and discussed for the entire SRC family of classifiers.

In Part I of this paper, an operational fully automated Landsat-like image spectral rule-based decision-tree classifier (LSRC), suitable for mapping radiometrically calibrated seven-band Landsat-4/-5 Thematic Mapper (TM) and Landsat-7 Enhanced TM+ (ETM+) spaceborne images [eventually synthesized from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) and the Moderate Resolution Imaging Spectroradiometer (MODIS) imaging sensor] into a discrete and finite set of spectral categories, has been downscaled to properly deal with spaceborne multispectral imaging sensors whose spectral resolution overlaps with, but is inferior to Landsat's, namely: 1) Satellite Pour l'Observation de la Terre (SPOT)-4/-5, Indian Remote Sensing Satellite (IRS)-1C/-1D/-P6 Linear Imaging Self-Scanner (LISS)-III, and IRS-P6 Advanced Wide Field Sensor (AWiFS); 2) National Oceanic and Atmospheric Administration (NOAA) Advanced Very High Resolution Radiometer (AVHRR) and Meteosat Second Generation (MSG); 3) Environmental Satellite (ENVISAT) Advanced Along-Track Scanning Radiometer (AATSR); 4) GeoEye-1, IKONOS-2, QuickBird-2, OrbView-3, TopSat, KOrean MultiPurpose SATellite (KOMPSAT)-2, FORMOsa SATellite (FORMOSAT)-2, Advanced Land Observing Satellite (ALOS) Advanced Visible and Near Infrared Radiometer type 2 (AVNIR-2), RapidEye, WorldView-2, PLEIADES-1/-2, and SPOT-6/-7; and 5) Disaster Monitoring Constellation (DMC), IRS-P6 LISS-IV, and SPOT-1/-2. LSRC, together with its five downscaled versions, identified, respectively, as the four-band SPOT-like SRC (SSRC), the four-band AVHRR-like SRC (AVSRC), the five-band AATSR-like SRC (AASRC), the four-band IKONOS-like SRC (ISRC), and the three-band DMC-like SRC (DSRC), form the so-called integrated SRC system of systems. In this paper, first, the classification accuracy and robustness to changes in the input data set of SSRC, AVSRC, AASRC, ISRC, and DSRC are assessed, both qualitatively and quantitatively, in comparison with LSRC's. Next, ongoing and future SRC applications are presented and discussed. They encompass: 1) the implementation of operational two-stage stratified hierarchical Remote Sensing (RS) image understanding systems discussed in Part I of this paper; 2) the integration of near real-time satellite mapping services with Internet map servers; and 3) the development of a new approach to semantic querying of large-scale multisensor image databases. These experimental results and application examples prove that the integrated SRC system of systems is operational, namely, it is effective, near real-time, automatic, and robust to changes in the input data set. Therefore, SRC appears eligible for use in operational satellite-based measurement systems such as those envisaged by the ongoing international Global Earth Observation System of Systems (GEOSS) Programme and the Global Monitoring for Environment and Security (GMES) system project.

In the above titled two-part paper (ibid., vol. 48, no. 3, pp. 1299-1325 and 1326-1354, Mar. 2010), there are errors which are corrected here.

The increasing amount of remote sensing (RS) imagery acquired from multiple platforms and the recent announcements that scientists and decision makers around the world will soon have unrestricted access at no charge to large-scale spaceborne multispectral (MS) image databases make urgent the need to develop easy-to-use, effective, efficient, robust, and scalable satellite-based measurement systems. In these scientific and industrial contexts, it is well known that, to date, the operational performance of existing stratified non-Lambertian (anisotropic) topographic correction (SNLTOC) algorithms has been limited by the need fora prioriknowledge of structural landscape characteristics, such as surface roughness which is land cover class specific. In practice, to overcome the circular nature of the SNLTOC problem, a mutually exclusive and totally exhaustive land cover classification map of a spaceborne MS image is requiredbeforeSNLTOC takes place. This system requirement is fulfilled by the original operational automatic two-stage SNLTOC approach presented in this paper which comprises, in cascade, 1) an automatic stratification first stage and 2) a second-stage ordinary SNLTOC method selected from the literature. The former combines 1) four subsymbolic digital-elevation-model-derived strata, namely, horizontal areas, self-shadows, and sunlit slopes either facing the sun or facing away from the sun, and 2) symbolic (semantic) strata generated from the input MS image by an operational fully automated spectral-rule-based decision-tree preliminary classifier recently presented in RS literature. In this paper, first, previous works related to the TOC subject are surveyed, and next, the novel operational two-stage SNLTOC system is presented. Finally, the original two-stage SNLTOC system is validated in up to 19 experiments where the system's capability of reducing within-stratum spectral variance while preserving pixel-based spectral patterns (shapes) is assessed quantitatively.

In the last 20 years, the number of spaceborne very high resolution (VHR) optical imaging sensors and the use of satellite VHR optical images have continued to increase both in terms of quantity and quality of data. This has driven the need for automating quantitative analysis of spaceborne VHR optical imagery. Unfortunately, existing remote sensing image understanding systems (RS-IUSs) score poorly in operational contexts. In recent years, to overcome operational drawbacks of existing RS-IUSs, an original two-stage stratified hierarchical RS-IUS architecture has been proposed by Shackelford and Davis. More recently, an operational automatic pixel-based near-real-time four-band IKONOS-like spectral rule-based decision-tree classifier (ISRC) has been downscaled from an original seven-band Landsat-like SRC (LSRC). The following is true for ISRC: (1) It is suitable for mapping spaceborne VHR optical imagery radiometrically calibrated into top-of-atmosphere or surface reflectance values, and (2) it is eligible for use as the pixel-based preliminary classification first stage of a Shackelford and Davis two-stage stratified hierarchical RS-IUS architecture. Given the ISRC “full degree” of automation, which cannot be surpassed, and ISRC computation time, which is near real time, this paper provides a quantitative assessment of ISRC accuracy and robustness to changes in the input data set consisting of 14 multisource spaceborne images of agricultural landscapes selected across the European Union. The collected experimental results show that, first, in a dichotomous vegetation/nonvegetation classification of four synthesized VHR images at regional scale, ISRC, in comparison with LSRC, provides a vegetation detection accuracy ranging from 76% to 97%, rising to about 99% if pixels featuring a low leaf area index are not considered in the comparison. Second, in the generation of a binary vegetation mask from ten panchromatic-sharpened QuickBird-2 and IKONOS-2 images, the operational performance measurement of ISRC is superior to that of an ordinary normalized difference vegetation index thresholding technique. Finally, the second-stage automatic stratified texture-based separation of low-texture annual cropland or herbaceous range land (land cover class AC/HR) from high-texture forest or woodland (land cover class F/W) is performed in the discrete, finite, and symbolic ISRC map domain in place of the ordinary continuous varying, subsymbolic, and multichannel texture feature domain. To conclude, this paper demonstrates that the automatic ISRC is eligible for use in operational VHR satellite-based measurement systems such as those envisaged under the ongoing Global Earth Observation System of Systems (GEOSS) and Global Monitoring for the Environment and Security (GMES) international programs.

The variogram image extracted from the panchromatic band of a QuickBird image has been analyzed to discriminate between similar spectral vegetated land use classes, such as forest and dense olive groves, once spectral features have been exploited. Such images seem to reproduce the direction and inter-row spacing of oriented texton-based textures present in the original input image. The variogram based texture algorithm (VBTA), recently proposed in literature for analyzing aerial images, tries to determine if texture in a region is not oriented, as in the case of forested and grassland areas or is oriented in one or two main directions, as in the case of vineyards and olive groves, respectively. If texture is oriented, the number of orientations, the orientation angle and periods (i.e. the distance between textons in each direction) are extracted, the latter depending on agricultural practices. This paper concentrates on: 1) the experimental selection of some variables useful for variogram analysis: i) the size of the input image moving window for variogram calculation; ii) some threshold values to be used for variogram normalization and for testing the presence/absence of any directionality in the variogram image; 2) The methodology for orientation and period estimation from the cumulative variance of the variogram image. The results obtained for a Southern Italy area characterized by intensive olive trees cultivation reveals the effectiveness of VBTA for QuickBird images.

Although underestimated in practice, the small/unrepresentative sample problem is likely to affect a large segment of real-world remotely sensed (RS) image mapping applications where ground truth knowledge is typically expensive, tedious, or difficult to gather. Starting from this realistic assumption, subjective (weak) but ample evidence of the relative effectiveness of existing unsupervised and supervised data labeling systems is collected in two RS image classification problems. To provide a fair assessment of competing techniques, first the two selected image datasets feature different degrees of image fragmentation and range from poorly to ill-posed. Second, different initialization strategies are tested to pass on to the mapping system at hand the maximally informative representation of prior (ground truth) knowledge. For estimating and comparing the competing systems in terms of learning ability, generalization capability, and computational efficiency when little prior knowledge is available, the recently published data-driven map quality assessment (DAMA) strategy, which is capable of capturing genuine, but small, image details in multiple reference cluster maps, is adopted in combination with a traditional resubstitution method. Collected quantitative results yield conclusions about the potential utility of the alternative techniques that appear to be realistic and useful in practice, in line with theoretical expectations and the qualitative assessment of mapping results by expert photointerpreters.

Based on purely spectral-domain prior knowledge taken from the remote sensing (RS) literature, an original spectral (fuzzy) rule-based per-pixel classifier is proposed. Requiring no training and supervision to run, the proposed spectral rule-based system is suitable for the preliminary classification (primal sketch, in the Marr sense) of Landsat-5 Thematic Mapper and Landsat-7 Enhanced Thematic Mapper Plus images calibrated into planetary reflectance (albedo) and at-satellite temperature. The classification system consists of a modular hierarchical top-down processing structure, which is adaptive to image statistics, computationally efficient, and easy to modify, augment, or scale to other sensors' spectral properties, like those of the Advanced Spaceborne Thermal Emission and Reflection Radiometer and of the Satellite Pour l'Observation de la Terre (SPOT-4 and -5). As output, the proposed system detects a set of meaningful and reliable fuzzy spectral layers (strata) consistent (in terms of one-to-one or many-to-one relationships) with land cover classes found in levels I and II of the U.S. Geological Survey classification scheme. Although kernel spectral categories (e.g., strong vegetation) are detected without requiring any reference sample, their symbolic meaning is intermediate between those (low) of clusters and segments and those (high) of land cover classes (e.g., forest). This means that the application domain of the kernel spectral strata is by no means alternative to RS data clustering, image segmentation, and land cover classification. Rather, prior knowledge-based kernel spectral categories are naturally suitable for driving stratified application-specific classification, clustering, or segmentation of RS imagery that could involve training and supervision. The efficacy and robustness of the proposed rule-based system are tested in two operational RS image classification problems

This work presents a design-based validation and calibration scheme for the Global Burned Area 2000 (GBA2000) products. The objective of such a scheme is to assess the margins of uncertainty associated with the burned area products and to estimate calibration coefficients needed to convert burned pixel counts into areal estimates. As the validation of GBA2000 was performed long after 2000, and given the fact that burned areas are a predominantly nonpermanent land cover change, the reference data are obtained from a set of Landsat-7 Enhanced Thematic Mapper Plus high-resolution remotely sensed data. A stratified sampling scheme is presented, specifically designed for the retrospective validation of burned area data; the scheme is based on combining information from two low-resolution burned area products (GBA2000 itself and Globscar). The resulting stratification has been applied to the whole global GBA2000 dataset, and preliminary validation results are reported for Africa. The conclusions highlight the limits of a retrospective validation exercise, and summarize some of the open issues in the validation of global burned area maps

This paper deals with the problem of badly posed image classification. Although underestimated in practice, bad-posedness is likely to affect many real-world image classification tasks, where reference samples are difficult to collect (e.g., in remote sensing (RS) image mapping) and/or spatial autocorrelation is relevant. In an image classification context affected by a lack of reference samples, an original inductive learning multiscale image classifier, termed multiscale semisupervised expectation maximization (MSEM), is proposed. The rationale behind MSEM is to combine useful complementary properties of two alternative data mapping procedures recently published outside of image processing literature, namely, the multiscale modified Pappas adaptive clustering (MPAC) algorithm and the sample-based semisupervised expectation maximization (SEM) classifier. To demonstrate its potential utility, MSEM is compared against nonstandard classifiers, such as MPAC, SEM and the single-scale contextual SEM (CSEM) classifier, besides against well-known standard classifiers in two RS image classification problems featuring few reference samples and modestly useful texture information. These experiments yield weak (subjective) but numerous quantitative map quality indexes that are consistent with both theoretical considerations and qualitative evaluations by expert photointerpreters. According to these quantitative results, MSEM is competitive in terms of overall image mapping performance at the cost of a computational overhead three to six times superior to that of its most interesting rival, SEM. More in general, our experiments confirm that, even if they rely on heavy class-conditional normal distribution assumptions that may not be true in many real-world problems (e.g., in highly textured images), semisupervised classifiers based on the iterative expectation maximization Gaussian mixture model solution can be very powerful in practice when: 1) there is a lack of reference samples with respect to the problem/model complexity and 2) texture information is considered negligible (i.e., a piecewise constant image model holds).

This work focuses on two challenging types of problems related to quality assessment and comparison of thematic maps generated from remote sensing (RS) images when little or no ground truth knowledge is available. These problems occur when: (1) competing thematic maps, generated from the same input RS image, assumed to be available, must be compared, but no ground truth knowledge is found to assess the accuracy of the mapping problem at hand, and (2) the generalization capability of competing classifiers must be estimated and compared when the small/unrepresentative ground truth problem affects the RS inductive learning application at hand. Specifically focused on badly posed image classification tasks, this paper presents an original data-driven (i.e., unsupervised) thematic map quality assessment (DAMA) strategy complementary (not alternative) in nature to traditional supervised map accuracy assessment techniques, driven by the expensive and error-prone digitization of ground truth knowledge. To compensate for the lack of supervised regions of interest, DAMA generates so-called multiple reference cluster maps from several blocks of the input RS image that are clustered separately. Due to the unsupervised (i.e., subjective) nature (ill-posedness) of data clustering, DAMA provides no (absolute) map accuracy measure. Rather, DAMA's map quality indexes are to be considered unsupervised (i.e., subjective) relative estimates of labeling and segmentation consistency between every competing map at hand and the set of multiple reference cluster maps. In two badly posed RS image mapping experiments, DAMA's map quality measures are proven to be: (1) useful in the relative comparison of competing mapping systems; (2) consistent with theoretical expectations; and (3) in line with mapping quality criteria adopted by expert photointerpreters. Documented limitations of DAMA are that it is intrinsically heuristic due to the subjective nature of the clustering problem, and like any evaluation measure, it cannot be injective.

The Global Rain Forest Mapping (GRFM) radar mosaics, generated from L-band Japanese Earth Resources Satellite 1 imagery downsampled to 100-m pixel size, provide a two-season spatially continuous coverage of the humid tropical ecosystems of the world. This paper presents a novel classification approach suitable for regional-scale vegetation mapping using the GRFM datasets. The mapping system consists of: 1) an application-dependent wavelet-based edge-preserving smoothing algorithm and 2) a two-stage per-pixel hybrid learning nearest multiple-prototype (NMP) classifier, whose unsupervised first stage is a per-pixel near-optimal vector quantizer, called enhanced Linde-Buzo-Gray (ELBG), recently proposed in pattern recognition literature. Identified as ENMP (NMP with ELBG), this novel classification approach is compared against two alternative systems in the classification of forest cover disturbances located across an area in the Amazon Basin. Surface classes of interest are primary forest, degraded forest, nonforest, and water bodies. Reference maps, derived from 30-m resolution Landsat Thematic Mapper imagery, are provided by the National Aeronautics and Space Administration and the Food and Agriculture Organization of the United Nations. Abundant quantitative and qualitative evidence shows that: 1) in a forest/nonforest data-mapping task, ENMP provides a testing accuracy of 87%, in line with training accuracies, i.e., the proposed method seems capable of generalizing well over the GRFM South America dataset and 2) among three competing approaches, ENMP provides the best compromise between ease of use, mapping accuracy, and computational time. Starting from these results, ENMP is employed to generate a swamp forest map of the whole Amazon Basin from the two-season GRFM radar mosaic of South America, in the context of the Global Land Cover project (GLC 2000).

The Global Rain Forest Mapping project (GRFM) is an initiative started by the National Agency for Space Development of Japan (NASDA) in 1996 with the main goal of creating a wall to wall radar map of the tropical belt with homogeneous and consistent characteristics. GRFM Africa-the part of the project related to tropical Africa-has evolved through several years to the stage where significant thematic products have been generated. It is maintained that these products bear relevance to global change studies and to the sustainable management of local resources in the tropics. The objective of this paper is to lend support to this proposition by illustrating through a few examples the results achieved so far. In particular two land cover maps are presented covering respectively the Central Congo basin, and the Gabon country. Validation of these large-scale high-resolution products poses a challenging problem. The method adopted in GRFM Africa is outlined. It is based on comparison with independent thematic information with known error budget, derived from a combination of optical remote sensing observations, national forestry maps and ground surveys.

The modified adaptive pappas clustering (MPAC) algorithm, previously published in the image processing literature, is proposed as a valuable tool in the analysis of remotely sensed images where texture information is negligible. Owing to its contextual, adaptive, and multiresolutional labeling approach, MPAC preserves genuine but small regions, is easy to use (i.e., it requires minor user interaction to run), and is robust to changes in input parameters. As an application example, an MPAC-based three-stage classifier is applied to degraded forest detection in Landsat Thematic Mapper (TM) scenes of the Brazilian Amazon, where intermediate states of forest alterations caused by anthropogenic activities can be characterized by image structures 1-3 pixels wide. In three TM images of the Para test site, where classification results are validated by means of qualitative and quantitative comparisons with aerial photos, degraded forest areas cover 13% to 45% of the image ground coverage. In the Mato Grosso test site, the degraded forest class overlaps with 1) 10% of the closed-canopy forest detected by the deforestation mapping program of the Food and Agriculture Organization (FAO, 1992), and 2) 19% of the closed-canopy forest detected by the Tropical Rain Forest Information Center (TRFIC, 1996). These figures are in line with the conclusions of a study where present estimates of annual deforestation for the Brazilian Amazon are speculated to capture less than half of the forest area that is actually impoverished each year.

For pt.I see ibid., p.645-61 (2002). Part I of this paper defines the class of constructive unsupervised on-line learning simplified adaptive resonance theory (SART) clustering networks. Proposed instances of class SART are the symmetric fuzzy ART (S-Fuzzy ART) and the Gaussian ART (GART) network. In Part II of our work, a third network belonging to class SART, termed fully self-organizing SART (FOSART), is presented and discussed. FOSART is a constructive, soft-to-hard competitive, topology-preserving, minimum-distance-to-means clustering algorithm capable of: 1) generating processing units and lateral connections on an example-driven basis and 2) removing processing units and lateral connections on a minibatch basis. FOSART is compared with Fuzzy ART, S-Fuzzy ART, GART and other well-known clustering techniques (e.g., neural gas and self-organizing map) in several unsupervised learning tasks, such as vector quantization, perceptual grouping and 3-D surface reconstruction. These experiments prove that when compared with other unsupervised learning networks, FOSART provides an interesting balance between easy user interaction, performance accuracy, efficiency, robustness, and flexibility.

For pt.I see ibid., p.645-61 (2002). Part I of this paper defines the class of constructive unsupervised on-line learning simplified adaptive resonance theory (SART) clustering networks. Proposed instances of class SART are the symmetric fuzzy ART (S-Fuzzy ART) and the Gaussian ART (GART) network. In Part II of our work, a third network belonging to class SART, termed fully self-organizing SART (FOSART), is presented and discussed. FOSART is a constructive, soft-to-hard competitive, topology-preserving, minimum-distance-to-means clustering algorithm capable of: 1) generating processing units and lateral connections on an example-driven basis and 2) removing processing units and lateral connections on a minibatch basis. FOSART is compared with Fuzzy ART, S-Fuzzy ART, GART and other well-known clustering techniques (e.g., neural gas and self-organizing map) in several unsupervised learning tasks, such as vector quantization, perceptual grouping and 3-D surface reconstruction. These experiments prove that when compared with other unsupervised learning networks, FOSART provides an interesting balance between easy user interaction, performance accuracy, efficiency, robustness, and flexibility.

Mixed pixels are a major source of inconvenience in the classification of remotely sensed data. This paper compares MLP with so-called neuro-fuzzy algorithms in the estimation of pixel component cover classes. Two neuro-fuzzy networks are selected from the literature as representatives of soft classifiers featuring different combinations of fuzzy set-theoretic principles with neural network learning mechanisms. These networks are: 1) the fuzzy multilayer perceptron (FMLP) and 2) a two-stage hybrid (TSH) learning neural network whose unsupervised first stage consists of the fully self-organizing simplified adaptive resonance theory (FOSART) clustering model, FMLP, TSH, and MLP are compared on CLASSITEST, a standard set of synthetic images where per-pixel proportions of cover class mixtures are known a priori. Results are assessed by means of evaluation tools specifically developed for the comparison of soft classifiers. Experimental results show that classification accuracies of FMLP and TSH are comparable, whereas TSH is faster to train than FMLP. On the other hand, FMLP and TSW outperform MLP when little prior knowledge is available for training the network, i.e., when no fuzzy training sites, describing intermediate label assignments, are available.

Clustering algorithms aim at modeling fuzzy (i.e., ambiguous) unlabeled patterns efficiently. Our goal is to propose a theoretical framework where the expressive power of clustering systems can be compared on the basis of a meaningful set of common functional features. Part I of this paper reviews the following issues related to clustering approaches found in the literature: relative (probabilistic) and absolute (possibilistic) fuzzy membership functions and their relationships to the Bayes rule, batch and on-line learning, prototype editing schemes, growing and pruning networks, modular network architectures, topologically perfect mapping, ecological nets and neuro-fuzziness. From this discussion an equivalence between the concepts of fuzzy clustering and soft competitive learning in clustering algorithms is proposed as a unifying framework in the comparison of clustering systems. Moreover, a set of functional attributes is selected for use as dictionary entries in the comparison of clustering algorithms, which is the subject of part II of this paper.

For pt.I see ibid., p.775-85. In part I an equivalence between the concepts of fuzzy clustering and soft competitive learning in clustering algorithms is proposed on the basis of the existing literature. Moreover, a set of functional attributes is selected for use as dictionary entries in the comparison of clustering algorithms. In this paper, five clustering algorithms taken from the literature are reviewed, assessed and compared on the basis of the selected properties of interest. These clustering models are (1) self-organizing map (SOM); (2) fuzzy learning vector quantization (FLVQ); (3) fuzzy adaptive resonance theory (fuzzy ART); (4) growing neural gas (GNG); (5) fully self-organizing simplified adaptive resonance theory (FOSART). Although our theoretical comparison is fairly simple, it yields observations that may appear parodoxical. First, only FLVQ, fuzzy ART, and FOSART exploit concepts derived from fuzzy set theory (e.g., relative and/or absolute fuzzy membership functions). Secondly, only SOM, FLVQ, GNG, and FOSART employ soft competitive learning mechanisms, which are affected by asymptotic misbehaviors in the case of FLVQ, i.e., only SOM, GNG, and FOSART are considered effective fuzzy clustering algorithms.

Fuzzy learning vector quantization (FLVQ), also known as the fuzzy Kohonen clustering network, was developed to improve performance and usability of on-line hard-competitive Kohnen's vector quantization and soft-competitive self organizing map (SOM) algorithms. The FLVQ effectiveness seems to depend on the range of change of the weighting exponent m(t). In the first part of this work, extreme m(t) values (1 and 1, respectively) are employed to investigate FLVQ asymptotic behaviors. This analysis shows that when m(t) tends to either one of its extremes, FLVQ is affected by trivial vector quantization, which causes centroids collapse in the grand mean of the input data set. No analytical criterion has been found to improve the heuristic choice of the range of m(t) change. In the second part of this paper, two FLVQ and SOM classification experiments of remote sensed data are presented. In these experiments the two nets are connected in cascade to a supervised second stage, based on the delta rule. Experimental results confirm that FLVQ performance can be greatly affected by the user's definition of the range of change of the weighting exponent. Moreover, FLVQ shows instability when its traditional termination criterion is applied. Empirical recommendations are proposed for the enhancement of FLVQ robustness. Both the analytical and the experimental data reported seem to indicate that the choice of the range of m(t) change is still open to discussion and that alternative clustering neural-network approaches should be developed to pursue during training: 1) monotone reduction of the neurons' learning rate and 2) monotone reduction of the overlap among neuron receptive fields.

Prospective authors are requested to submit new, unpublished manuscripts for inclusion in the upcoming event described in this call for papers.

The class of the single linkage region growing (SLRG) algorithms, in which pairs of neighboring pixels are compared for merging, is one of the conceptually simplest approaches to image segmentation. A new normalized coefficient, which measures the degree of match between two multivalued vectors, termed the vector degree of match (VDM), provides the SLRG applications with the metric needed to group adjacent pixel pairs. Two new SLRG algorithms, applied to multiband images and exploiting the VDM criterion, are presented. Their major advantage, in comparison with the SLRG implementations found in the literature, is that they required only one user-defined parameter, the VDM Threshold (VDMT), which is a normalized value featuring local adaptivity and having an intuitive physical meaning. The use of an SLRG module is convenient as the first stage of a structured segmentation procedure featuring the following functional properties: (i) it is implemented sequentially; (ii) it combines single linkage, centroid linkage and hybrid linkage criteria; and (iii) its goal is the detection of image areas characterized by low contrast.

The aim of this study was to investigate the statistical meaning of six GLCM (Gray Level Cooccurrence Matrix) parameters. This objective was mainly pursued by means of a selfcorrsistent, theoretical assessment in order to remain independent from test image. The six statistical parameters are energy, contrast, variance, correlation, entropy and inverse dzfference moment, which are considered the most relevant among the 14 originally proposed by Haralick et al.. The functional analysis supporting theoretical considerations was based on natural clustering in the feature space of segment texture values. The results show that among the six GLCM statistical parameters, five different sets can be identified, each set featuring a specific textural meaning. The first set contains energy and entropy, while the four remaining parameters can be regarded as belonging to four different sets. Two parameters, energy and contrust, are considered to be the most efficient for discriminating different textural patterns. A new GLCM statistical parameter, recursivity, is presented in order to replace energy which presents some degree of correlation with contrast. It is demonstrated that in some cases it may be reasonable to replace the computation of GLCM with that of GLDH (Gray Level Difference Histogram), in order to benefit by a better compromise between texture measurement accuracy, computer storage and computation time.

This article presents an implementation of an artificial neural network (ANN) which performs unsupervised detection of recognition categories from arbitrary sequences of multivalued input patterns. The proposed ANN is called Simplified Adaptive Resonance Theory Neural Network (SARTNN). First, an Improved Adaptive Resonance Theory 1 (IARTl)-based neural network for binary pattern analysis is discussed and a Simplified ARTl (SART1) model is proposed. Second, the SARTl model is extended to multivalued input pattern clustering and SART” is presented. A normalized coefficient which measures the degree of match between two multivalued vectors, the Vector Degree of Match (VDM), provides SARTNN with the metric needed to perform clustering. Every ART architecture guarantees both plasticity and stability to the unsupervised learning stage. The SARTNN plasticity requirement is satisfied by implementing its attentional subsystem as a self-organized, feed-forward, flat Kohonen’s ANN (KANN). The SARTNN stability requirement is properly driven by its orienting subsystem. SARTNN processes multivalued input vectors while featuring a simplified architectural and mathematical model with respect to both the ARTl and the AkT2 models, the latter being the ART model fitted to multivalued input pattern categorization. While the ART2 model exploits ten user-defined parameters, SARTNN requires only two user-defined parameters to be run: the first parameter is the vigilance threshold, p, that affects the network’s sensibility in detecting new output categories, whereas the second parameter, T, is related to the network's learning rate. Both parameters have an intuitive physical meaning and allow the user to choose easily the proper discriminating power of the category extraction algorithm. The SARTNN performance is tested as a satellite image clustering algorithm. A chromatic component extractor is recommended in a satellite image preprocessing stage, in order to pursue SARTNN invariant pattern recognition. In comparison with classical clustering algorithms like ISODATA, the implemented system gives good results in terms of ease of use, parameter robustness and computation time. SARTNN should improve the performance of a Constraint Satisfaction Neural Network (CSNN) for image segmentation. SARTNN, exploited as a self-organizing first layer, should also improve the performance of both the Counter Propagation Neural Network (CPNN) and the Reduced connectivity Coulomb Energy Neural Network (RCENN).

A modified version of the refined gamma maximum-a-posteriori (RGMAP) speckle filter, which is found in the literature, is presented. The traditional RGMAP speckle filter first defects contours belonging to step edges and thin linear structures, then applies the RGMAP filter to local statistics extracted from rectangular masks that do not cross image contours. The proposed modified RGMAP (MRGMAP) filter first exploits local operators belonging to the odd-symmetric filter category employed by RGMAP to detect image segments, then it computes local statistics over areas that are not necessarily rectangular, but are subsets of the image segments having any possible shape. Therefore, MRGMAP enhances the RGMAP ability in exploiting shape adaptive windowing near image contours, where speckle is not fully developed. The MRGMAP computation time is estimated to be of the same magnitude of that of the original RGMAP, the latter depending on the number of filter categories being employed. The qualitative and quantitative results of the MRGMAP filter applied to real SAR images are satisfactory as the filter seems to be effective in speckle removal whereas it retains edge sharpness and subtle details. However, tests on simulated SAR images must still be performed in order to provide definitive evidence supporting MRGMAP effectiveness. Since MRGMAP typically removes image structures featuring a constant reflectivity gradient, this filter is not particularly suitable for image enhancement in human photo-interpretation. MRGMAP can be rather employed as a preprocessing module in a computer-based SAR image classification procedure based on segment mean value analysis.>

A knowledge-based, hierarchical, unsupervised classification scheme for high-resolution multispectral satellite (HRMS) images is described. This scheme, which finds its conceptual bases in the work of Nagao and Matsuyama for structural analysis of aerial photographs, introduces a new filtering algorithm which is able to preserve fine linear structures of the image. An example of the application of this classification scheme to a Landsat Thematic Mapper multispectral image is presented.>