Classification Routines Research Articles

Sub-pixel land cover composition estimation using a linear mixture model and fuzzy membership functions

Abstract Mixed pixels occur commonly in remotely-sensed imagery, especially those with a coarse spatial resolution. They are a problem in land-cover mapping applications since image classification routines assume ‘pure’ or homogeneous pixels. By unmixing a pixel into its component parts it is possible to enableinter alia more accurate estimation of the areal extent of different land cover classes. In this paper two approaches to estimating sub-pixel land cover composition are investigated. One is a linear mixture model the other is a regression model based on fuzzy membership functions. For both approaches significant correlation coefficients, all >0·7, between the actual and predicted proportion of a land cover type within a pixel were obtained. Additionally a case study is presented in which the accuracy of the estimation of tropical forest extent is increased significantly through the use of sub-pixel estimates of land-cover composition rather than a conventional image classification.

Non-classificatory analysis and representation of heathland vegetation from remotely sensed imagery

Remotely sensed imagery are an attractive source of data for vegetation mapping. Conventional image classification routines used to produce thematic maps from remotely sensed imagery rely on a one-pixel-one-class approach and generate discrete thematic units. Where the environmental phenomena to be mapped exhibit gradients such a representation is inappropriate. This paper discusses the representation of semi-natural vegetation, drawing on examples of heathland vegetation that lie along continua. Two alternative approaches to the representation of heathland vegetation are examined, both of which aim to model the continuous character of the vegetation with measures of the strength of membership to ‘discrete classes’, namely probabilities of class membership from a maximum likelihood classification and fuzzy membership functions from the fuzzy c-means algorithm. Since both approaches imply partial class membership they can be considered as non-classificatory, even though the measures of the strength of class membership may be derived from classification routines. The measures of class membership generated from both approaches were found to be significantly correlated to the variations in heathland composition along a transect which graded from dry heath to wet heath/bog. Furthermore for cases drawn from the class end-points both approaches were able to discriminate class membership accurately.

Seafloor acoustic remote sensing with multibeam echo-sounders and bathymetric sidescan sonar systems

This paper examines the potential for remote classification of seafloor terrains using a combination of quantitative acoustic backscatter measurements and high resolution bathymetry derived from two classes of sonar systems currently used by the marine research community: multibeam echo-sounders and bathymetric sidescans sonar systems. The high-resolution bathymetry is important, not only to determine the topography of the area surveyed, but to provide accurate bottom slope corrections needed to convert the arrival angles of the seafloor echoes received by the sonars into true angles of incidence. An angular dependence of seafloor acoustic backscatter can then be derived for each region surveyed, making it possible to construct maps of acoustic backscattering strength in geographic coordinates over the areas of interest. Such maps, when combined with the high-resolution bathymetric maps normally compiled from the data output by the above sonar systems, could be very effective tools to quantify bottom types on a regional basis, and to develop automatic seafloor classification routines.

Visual impairment in Nordic children. I. Nordic registers and prevalence data.

A Nordic study group of ophthalmologists, NORDSYN, has compiled data from registers in Denmark, Finland, Iceland and Norway of 2527 visually impaired children. Each record contains the following information: sex, year of birth, year of registration, classification of visual impairment, ocular diagnosis, systemic diagnosis, aetiology and evt. additional impairments. The ocular diagnoses were compiled into groups, and coding systems for aetiology and additional impairment were developed. The sex distribution revealed a dominance of males compared to the general population at the same age. Cases with non-genetic aetiology showed--through to a lesser extent--the same relative preponderance of males. The diseases in males caused by x-linked genetic factors do, therefore, not fully explain the sex distribution observed in the study. The national prevalences for registration of childhood blindness (WHO-definition: best corrected visual acuity in the best eye less than 3/60 or visual field less than 10 degrees around fixation for the ages 0-15 years) are per 100,000 child-population aged 0-15 years: Denmark 41, Finland 15, Iceland 19 and Norway 15. The differences are primarily presumed to be due to varying efficiency in registration. The proportion of visually impaired children with an additional mobility, hearing or mental impairment is between one-third and one-half of the national materials, thus indicating the need for interdisciplinary tracing of and care for the visually impaired child. This study documents the need of uniform routines for data classification of visually impaired children. The quality of the data in the present study calls for caution in the interpretation of the prevalence estimates. Incidence studies are being prepared to obtain information on whether the amount and causes of visual impairment in children with or without multiple impairments are changing.

An automatic data reduction and transfer method to aid pattern recognition analysis and classification of NMR spectra.

A method of automatically generating reduced NMR data and transferring it between computers is proposed. These data can then be used as descriptors for input to non-parametric statistical routines for classification of the samples.

Weighted Standardization—A General Data Transformation Method Proceeding Classification Procedures

AbstractDuring preparatory steps of data for automatic classification routines, the amount of information contained by the character distribution is reduced by standardization of the character values. This information can be regained through special weighting schemes of standardized character values. Two weight coefficients derived from decomposition of the frequency distributions of the characters are proposed.

Lipids of aquatic sediments, recent and ancient

Computerized gas chromatography - mass spectrometry (g.c.-m.s.) is now an essential tool in the analysis of the complex mixtures of lipids (geolipids) encountered in aquatic sediments, both Recent (less than 1 Ma (10 6 years) old) and ancient. Most geolipid studies have been performed in the e.i. mode at low resolution but the techniques now being applied include c.i. and h.r.m.s. The large quantities of data acquired from capillary g.c.-m.s. runs necessitate fast data acquisition and data processing, including the capability for the automatic selection and refinement of key spectra. Even so, the chemist is faced with the identification and/or recognition of at least several hundred good quality spectra from a single run. Fast routine search procedures are useful here, especially for known compounds, while classification routines based on established rules for manual interpretation can be of assistance even with novel compounds. Examples from recent studies (at Bristol) of contemporary, Recent and ancient sediments, are presented. Geolipids show abundance patterns of homologous series which, while related to those of known organisms, display many novel features, including extensive carbon number ranges and stereospecific distributions. Additionally, certain carbon skeleton types, commonly thought to be rare as natural products, are major components of geolipid fractions, presumably reflecting the composite inputs and early microbial diagenetic activity.

Spectral signature selection for mapping unvegetated soils

Airborne multispectral scanner data covering the wavelength interval from 0.40–2.60 μm were collected at an altitude of 1000 m above the terrain in southeastern Pennsylvania. Uniform training areas were selected within three sites from this flightline. These training areas were field mapped as Berks, Duffield and Penn soils. Soil samples were collected from each site and a procedure developed to allow assignment of scan line and element number from the multispectral scanner data to each sampling location. These soil samples were analyzed on a Beckman DK-2A Spectrophotometer and laboratory spectral signatures derived. After correcting for solar radiation and atmospheric attenuation, these laboratory signatures were compared to the spectral signatures derived from these same soils using multispectral scanner data. Both signatures were used in supervised and unsupervised classification routines. Computer-generated maps using the laboratory and multispectral scanner derived signatures resulted in maps that were similar to maps resulting from field surveys. Approximately 90% agreement was obtained between classification maps produced using multispectral scanner derived signatures and laboratory derived signatures.

Nonlinear Regression With Linear Constraints: An Extension of the Magnified Diagonal Method

A syntax-directed picture analysis system based on a formal picture description scheme is described. The system accepts a description of a set of pictures in terms of a grammar generating strings in a picture description language; the grammar is explicitly used to direct the analysis or parse, and to control the calls on pattern classification routines for primitive picture components. Pictures are represented by directed graphs with labeled edges, where the edges denote elementary picture components and the graph connectivity mirrors the picture component connectivity; blank and don't care “patterns” allow the description of simple relations between visible patterns. The bulk of the paper is concerned with the picture parsing algorithm which is an n -dimensional analog of a classical top-down string parser, and an application of an implemented system to the analysis of spark chamber film. The potential benefits of this approach, as demonstrated by the application, include ease of implementation and modification of picture processing systems, and simplification of the pattern recognition problem by automatically taking advantage of contextual information.

Parsing of Graph-Representable Pictures

A syntax-directed picture analysis system based on a formal picture description scheme is described. The system accepts a description of a set of pictures in terms of a grammar generating strings in a picture description language; the grammar is explicitly used to direct the analysis or parse, and to control the calls on pattern classification routines for primitive picture components. Pictures are represented by directed graphs with labeled edges, where the edges denote elementary picture components and the graph connectivity mirrors the picture component connectivity; blank and don't care “patterns” allow the description of simple relations between visible patterns. The bulk of the paper is concerned with the picture parsing algorithm which is an n -dimensional analog of a classical top-down string parser, and an application of an implemented system to the analysis of spark chamber film. The potential benefits of this approach, as demonstrated by the application, include ease of implementation and modification of picture processing systems, and simplification of the pattern recognition problem by automatically taking advantage of contextual information.