A STATISTICAL APPROACH TO FEATURE CLASSIFICATION ON REMOTE SENSING IMAGERY

The mathematical basis for a feature classification algorithm is described which combines elements of game theory with Bayesian and suboptimal [feature classification] decision rules. Comparison of reflectance values with training area parameters, according to a sequence of diminishing a priori probabilities that the values will be assigned to that particular class results in reductions in computer time during classification. Results of the procedure are demonstrated through a pair of “before” and “after” images. Translated from: Metody kompleksnykh aerokosmicheskikh issledovaniy Sibiri, L. K. Zyat'kova, ed. Novosibirsk: Nauka, 1985, pp. 75–79.     

AN EXPERIMENT IN SYSTEMATIZING THE CARTOGRAPHIC IMAGES OF GEOMORPHOLOGICAL FEATURES

The authors attempt to identify and standardize the cartographic symbols (“images”) used to represent a variety of specific geomorphological features. A large number of cartographic representations of landform features found on geological-geomorphological maps are classified in terms of their basic pattern/configuration and the level of complexity within each major pattern type. It is argued that the pattern or configuration of a cartographic symbol used to represent a geomorphological feature should represent the spatial structure of that feature with sufficient reliability to serve as a key to an understanding of its genesis. Translated from: Geomorfologiya, 1987, No. 2, pp. 62-66.     

INTERVISIBLE FEATURES

Abstract

It is frequently required to find whether a feature A of height h ₀ will interrupt the view between two other features A₁ and A₂, of heights h ₁ and h ₂ respectively. Suppose that the right line from A₁ to A₂, whose zenith distance is ζ at A₁, has a height h at A; it is then obvious that no more is necessary than to compute h and compare it with the known height h ₀ of the feature A.     

NEW IDEAS IN A QUANTITATIVE APPROACH TO IMAGE PROCESSING

The authors present a statistical approach for multi-band image processing based on “synthesizing” processing methods, i.e., methods which transform N original band-specific images into a smaller number of (synthetic) image products. Such techniques are increasingly replacing “one-dimensional” methods (one-to-one correspondence between number of original and processed images) because of the normally greater informational content and enhanced feature discrimination capability afforded by composite image products. The ultimate objective is development of a complex of image processing strategies balancing user demands for information quantity, image diversity, and efficient use of computer time. Translated from: Metody kompleksnykh aerokosmicheskikh issledovaniy Sibiri, L. K. Zyat'kova, ed. Novosibirsk: Nauka, 1985, pp. 84-87.     

Area between Peaks Feature in the Derivative Reflectance Curve as a Sensitive Indicator of Change in Chlorophyll Concentration

Vegetation spectral features can detect chlorophyll concentrations. Two key spectral features evident in the first derivative (FD) of reflectance constitute the two main peaks: one located around 685-705 nm and the other near 710-725 nm. We propose that the area between peaks (ABP) can be used as a sensitive indicator of changes in the photosynthetic pigments at leaf level and demonstrate it using a high-spectral-resolution dataset of maize leaves collected by Gitelson and coworkers (2005). We find significant high positive correlations (r ² > 0.90) between chlorophyll concentrations and both the ABP and its continuum length feature.     

SELECTION OF AN OPTIMAL ALGORITHM FOR RECOGNITION OF NATURAL FEATURES

Two widely used Soviet algorithms for the classification of hydromete-orological features are compared for the purposes of mapping water temperature, electrical conductivity, pH, transparency/turbidity, dissolved surface concentration of chemical elements, depth, etc. The KLASS algorithm, using a solution to the “traveling salesman” problem for classifying objects in feature space, produced higher intra-class proximity and greater inter-class “distances” than the KSS algorithm, based on optimizing “shortest-path” criteria. This was determined through the use of five similar measures of recognition quality. Translated from: Vestnik Leningrad-skogo Universiteta, 1985, No. 14, pp. 95-100.     

A GEOGRAPHIC INFORMATION SYSTEM FOR PHYSICAL-GEOGRAPHIC REGIONALIZATION

The design and functioning of a geographic information system established for the purpose of objectively delineating the boundaries of major physical geographic regions in the USSR is described. The basic components include an information input block for data collection and preliminary image processing; a recognition block for feature classification and analysis; a data base management system providing for repeated revision, addition, and use of spectral information; and a block for displaying results of processing in various forms. Applications of the system are demonstrated in the mapping of physical regions in a steppe and semi-desert area of the southern USSR. Translated from: Izvestiya Akademii Nauk SSSR, seriya geograficheskaya, 1988, No. 2, pp. 89-94.     

DEVELOPMENT OF QUANTITATIVE METHODS OF STUDYING THE EARTH FROM SPACE

The paper surveys the development of “space geography”–a field of knowledge applying the methods of remote sensing, the physical sciences, and mathematics in the solution of geographic problems. Early advances featured methodological improvements (use of quantitative methods in image processing), whereas future research must focus upon perfecting our knowledge of: (a) relationships underlying the use of these methods, i.e., between environmental parameters and reflectance values, (b) methods of effectively combining different types of imagery, and image products and field work, in research, and (c) appropriate quantitative indices for feature recognition. Examples of the operationalization of such concerns are demonstrated for land-use and soil mapping projects. Translated from: Izvestiya, AN SSSR, seriya geograficheskaya, 1985, No. 5, pp. 110–116.     

REMOTE SENSING METHODS IN RESEARCH ON RIVER DRAINAGE BASINS

The concept that river drainage basins can be considered appropriate spatial units in assessment of human environmental impacts is examined. Applications of remote sensing imagery in the study of changes occurring within a particular basin—the Sit' River in European Russia—are investigated and multispectral aerial photography is used to determine the structure of biotic communities. Regionalization and areal cataloguing of these communities are the basis for specific conclusions regarding the principal sources of entry of nutrients and pollutants into the basin. Three image feature classification methods used to identify such communities (K-means (unsupervised) based on principal components analysis, supervised using principal components analysis, and data compression and subsequent identification of principal components) are compared in terms of effectiveness (numbers of mixed and/or unclassified pixels). Translated by Edward Torrey, Alexandria, VA 22308 from: Izvestiya Akademii Nauk, seriya geograficheskaya, 1994, No. 1, pp. 126-140.     

Enabling Spatially Aware Mobile Applications

Recently, increasing numbers of mobile phones are appearing on the market that feature advanced navigation capabilities: embedded GPS receivers for global positioning, integrated digital compasses for detecting the heading of the device, or accelerometer‐based tilt sensors will potentially enable upcoming and future mobile phones to measure their location and orientation in 3D space. In this paper, we present an application framework for building spatially aware mobile applications – applications that visualize, process or exchange geospatial information – on mobile phones equipped with such features. The core component of the framework is a novel, platform‐independent XML data exchange format for the interface between application server and mobile device that describes the geographic vicinity of the user. The format enables a variety of new mobile interaction styles and user interface types – from traditional text‐based local search and information interfaces to innovative real‐time user interfaces like Geo‐Wands and Smart Compasses.     

INTERPRETATION OF THE AGRICULTURAL LANDSCAPE WITH SPACE IMAGERY

Heterochronous multispectral imagery from the “Fragment” scanner is used to identify and map a series of “natural-agricultural complexes,” or agricultural landscapes, of southern European Russia. Interpretation is based on imagery in the green, orange, red, and near-infrared zones of the spectrum (0.5-1.1 μm). Interpretation keys and other information designed to facilitate feature discrimination (optimal wavelengths, best seasons for imagining) are provided in many instances. Natural landscape and soil erosion maps are also compiled from the imagery, which supply information (an optimal crop rotation scheme and needed reclamation measures) used on the agricultural landscape map. Translated from: Vestnik Moskovs-kogo Unlversiteta, geografiya, 1985, No. 2, pp. 34-41.     

THE LANGUAGE OF MAPS AND ITS PRINCIPAL FEATURES

The author, head of the Department of Cartography at the Institute of Geography, analyzes the language of maps in terms of general semiotic principles. Using the similarity of processes in other language systems, he constructs a model of the language of maps that distinguishes two sublanguages, one reflecting the locational aspects of maps (coordinate grids, etc.), the other reflecting the map content. The discussion touches on the system of map symbols as a linguistic feature, analyzes the system of map language in terms of such linguistic features as vocabulary, grammar and syntax, and draws other analogies between the language of maps and other language systems. Translated from: Voprosy Geografii, No. 122/123, 1984, pp. 40-56.     

Noise-signal index threshold: a new noise-reduction technique for generation of reference spectra and efficient hyperspectral image classification

Reference spectra of terrestrial targets are usually collected using field spectro-radiometers for mineral abundance mapping and target detection. These spectra often have noise that masks characteristic absorption and reflection features and affects the efficiency of material mapping. This work aims at obtaining an empirical technique for reduction of high-frequency noise from field spectra. The proposed noise correction technique uses a ‘normalized’ measure R_n , where R_n  = (L_n  ? F_n )/L_n for each band (n) calculated from field and laboratory spectra of test material, with F_n and L_n being the depth of the absorption feature in field and laboratory spectra, respectively. On the basis of the assumption of the constancy of this ratio in neighbouring bands, an empirical algorithm that approximates the ratio R_n of a noisy band to the corrected ratio of an adjacent band is used to obtain the noise-corrected field spectra. The classification accuracy increases significantly when noise reduced field spectra are used as reference spectra.     

Mapping grassland leaf area index with airborne hyperspectral imagery: A comparison study of statistical approaches and inversion of radiative transfer models

Statistical and physical models have seldom been compared in studying grasslands. In this paper, both modeling approaches are investigated for mapping leaf area index (LAI) in a Mediterranean grassland (Majella National Park, Italy) using HyMap airborne hyperspectral images. We compared inversion of the PROSAIL radiative transfer model with narrow band vegetation indices (NDVI-like and SAVI2-like) and partial least squares regression (PLS). To assess the performance of the investigated models, the normalized RMSE (nRMSE) and R² between in situ measurements of leaf area index and estimated parameter values are reported. The results of the study demonstrate that LAI can be estimated through PROSAIL inversion with accuracies comparable to those of statistical approaches (R² = 0.89, nRMSE = 0.22). The accuracy of the radiative transfer model inversion was further increased by using only a spectral subset of the data (R² = 0.91, nRMSE = 0.18). For the feature selection wavebands not well simulated by PROSAIL were sequentially discarded until all bands fulfilled the imposed accuracy requirements.     

TLS-bridged co-prediction of tree-level multifarious stem structure variables from worldview-2 panchromatic imagery: a case study of the boreal forest

ABSTRACT

In forest ecosystem studies, tree stem structure variables (SSVs) proved to be an essential kind of parameters, and now simultaneously deriving SSVs of as many kinds as possible at large scales is preferred for enhancing the frontier studies on marcoecosystem ecology and global carbon cycle. For this newly emerging task, satellite imagery such as WorldView-2 panchromatic images (WPIs) is used as a potential solution for co-prediction of tree-level multifarious SSVs, with static terrestrial laser scanning (TLS) assumed as a ‘bridge’. The specific operation is to pursue the allometric relationships between TLS-derived SSVs and WPI-derived feature parameters, and regression analyses with one or multiple explanatory variables are applied to deduce the prediction models (termed as Model1s and Model2s). In the case of Picea abies, Pinus sylvestris, Populus tremul and Quercus robur in a boreal forest, tests showed that Model1s and Model2s for different tree species can be derived (e.g. the maximum R²?=?0.574 for Q. robur). Overall, this study basically validated the algorithm proposed for co-prediction of multifarious SSVs, and the contribution is equivalent to developing a viable solution for SSV-estimation upscaling, which is useful for large-scale investigations of forest understory, macroecosystem ecology, global vegetation dynamics and global carbon cycle.     

Multiple statistical approaches for the discrimination of mangrove species of Rhizophoraceae using transformed field and laboratory hyperspectral data

This study aims at discriminating eight mangrove species of Rhizophoraceae family of Indian east coast using field and laboratory spectra in spectral range (350–2500 nm). Parametric and non-parametric statistical analyses were applied on spectral data in four spectral modes: (i) reflectance (ii) continuum removed, (iii) additive inverse and (iv) continuum removed additive inverse. We introduced continuum removal of inverse spectra to utilize the advantage of continuum removal in reflectance region. Non-parametric test gave better separability than parametric test. Principal component analysis and stepwise discriminant analysis were applied for feature reduction and to identify optimal wavelengths for species discrimination. To quantify the separability, Jeffries–Matusita distance measure was derived. Green (550 nm), red edge (680–720 nm) and water absorption region (1470 and 1850 nm) were found to be optimal wavelengths for species discrimination. The continuum removal of additive inverse spectra gave better separability than the continuum removed spectra.     

Pattern detection in airborne LiDAR data using Laplacian of Gaussian filter

Methods for feature detection in laser scanning data have been studied for decades ever since the emergence of the technology.However,it is still one of the unsolved problems in LiDAR data processing due to difficulty of texture and structure information extraction in unevenly sampled points.The paper analyzes the characteristics of Laplacian of Gaussian(LoG) Filter and its potential use for structure detection in LiDAR data.A feature detection method based on LoG filtering is presented and ex-perimented on the unstructured points.The method filters the elevation value(namely,z coordinate value) of each point by convo-lution using LoG kernel within its local area and derives patterns suggesting the existence of certain types of ground ob-jects/features.The experiments are carried on a point cloud dataset acquired from a neighborhood area.The results demonstrate patterns detected at different scales and the relationship between standard deviation that defines LoG kernel and neighborhood size,which specifies the local area that is analyzed.     

Comparative study on projected clustering methods for hyperspectral imagery classification

In this study, projected clustering is introduced to hyperspectral imagery for unsupervised classification. The main advantage of projected clustering lies in its ability to simultaneously perform feature selection and clustering. This framework also allows selection of different sets of dimensions (features/bands) for different clusters. This framework provides an effective way to address the issues associated with the high dimensionality of the data. Experiments are conducted on both synthetic and real hyperspectral imagery. For this purpose, projected clustering algorithms are implemented and compared with k-means and k-means preceded by principal component analysis. Preliminary analyses of studied algorithms on synthetic hyperspectral imagery demonstrate good results. For real hyperspectral imagery, only ORCLUS is able to produce acceptable results as compared to other unsupervised methods. The main concern lies with identification of right parameter settings. More experiments are required in this direction.