Classification Of Satellite Imagery Research Articles

A comparison of land cover types in an ecological field survey in Northern England and a remotely sensed land cover map of Great Britain

Earth-orbiting remote sensing satellites have important roles to play in resource assessment and monitoring changes in land cover. However, there will often be a need to relate the land cover types identified from the classification of satellite imagery to other classifications of land cover, and particularly those familiar to practical conservationists working in the field. This paper compares the land cover types recognised in the first remotely sensed land cover map of Great Britain with those of a widely used ecological field survey method (the Phase-1 survey technique of the former Nature Conservancy Council of Great Britain) within a sample area of 188 km 2 in Northern England. Digitised field maps were combined with the remotely sensed land cover map within a geographical information system (GIS) to produce a matrix of correspondence linking the two classifications. The results are discussed with an emphasis on the factors underlying the observed differences between the two sets of map data. Two potential applications of the land cover map are investigated — first in the mapping of dwarf shrub vegetation (a habitat which has declined over the last 50 years), and second in the detection of coniferous afforestation (a significant cause of semi-natural habitat loss in upland Britain).

Nonlinear aspects of data integration for land-cover classification in a neural network environment

Some results concerning the exploration of a neural network methodology's nonlinear aspects to provide land-cover in satellite imagery are presented. All required images are used in a Back-Error Propagation (BEP) network which is a nonlinear data integrator for spatial patterns classification. The network is trained to recognize the basic categories: grass, moisted soil, bare soil, forest, water and built-up areas. The results of a partial classification are used in a posterior analysis which is made to get the final classification in more detailed classes of land use. The performance results show how powerful is a neural-network based methodology for satellite imagery integration and classification.

An inexpensive and effective basis for monitoring rice areas using GIS and remote sensing

To ensure orderly, planned rice production, and to meet environmental constraints on the location of rice in the landscape, there is a need to locate and measure the area of rice plantings by paddock and by farm. Traditionally, rice areas have been estimated from manual measurements taken from aerial photography: overall, an expensive and time-consuming approach. In a pilot study covering 5 seasons on an area of 6 by 6.25 km containing 160 paddocks in 22 farms, we have found that it is possible to locate rice areas accurately, to determine their size, and to classify the vigour of the crop. Information is transferred into a geographic information system from classified satellite imagery using Landsat Thematic Mapper bands 3, 4, and 5 and then processed with a simple rule to distinguish rice rapidly from other crops. Furthermore, the classification categorises rice within each paddock into 3 statistically distinct groups of poor, medium, or good crop vigour. Total rice area and area of each rice class are computed for each rice paddock. Paddock data are combined to give corresponding values for each farm and for the entire study area. This information could be useful for forecasting rice yields in the current season and for the management of paddocks for more uniform and higher yields in subsequent seasons. The procedures used require inexpensive software and are largely automated since they include unsupervised clustering. The need for special skills is thereby minimised, making the transfer of the technology to interested parties quite straightforward. Over the 5 seasons, all paddocks growing rice were correctly identified and only 2 non-rice paddocks wrongly identified as growing rice.

Use of Geographic Information Systems for Applications on Gas Pipeline Rights-of-Way: ABSTRACT

Geographic information system (GIS) applications for the siting and monitoring of gas pipeline rights-of-way (ROWS) were developed for areas near Rio Vista, California. The data layers developed for this project represent geographic features, such as landcover, elevation, aspect, slope, soils, hydrography, transportation, endangered species, wetlands, and public line surveys. A GIS was used to develop and store spatial data from several sources; to manipulate spatial data to evaluate environmental and engineering issues associated with the siting, permitting, construction, maintenance, and monitoring of gas pipeline ROWS; and to graphically display analysis results. Examples of these applications include (1) determination of environmentally sensitive areas, such as endangered species habitat, wetlands, and areas of highly erosive soils; (2) evaluation of engineering constraints, including shallow depth to bedrock, major hydrographic features, and shallow water table; (3) classification of satellite imagery for land use/landcover that will affect ROWS; and (4) identification of alternative ROW corridors that avoid environmentally sensitive areas or areas with severe engineering constraints.

Remote sensing of soil processes

Estimates of ecosystem structure and function using remotely sensed data are based on the information contained in reflected and emitted radiation. Soil plays an integral role in almost all remote observations of the land surface either because the vegetation cover is discontinuous or because multiple scattering of the radiation occurs between vegetation and the underlying soil surface. However, since vegetation covers most of the land surface, we should consider the plant canopy as a primary interface to soil processes. In many ways, the canopy is an integrated expression of (and is mutually linked to) below-ground processes and can, as such, be used to infer ecosystem properties. Classifications of satellite imagery into vegetation communities in which soil processes have been characterized can be used to extrapolate those processes across the landscape based on the areal extent of the associated cover type. Ratios of reflectance values measured in the region of chlorophyll absorption to those in the highly reflected near-infrared region are sensitive to biomass quantity and, when integrated over time, strongly correlate with primary productivity. Imaging spectrometers which measure the reflected spectrum in many narrow contiguous bands enable the development of spectral mixture models to deconvolve the reflectance signal from each pixel in an image into each of the components contributed by spectrally distinct materials on the ground, e.g. vegetation and soil. High spectral resolution measurements of the absorption characteristics of major foliar chemical constituents may permit assessment of canopy-level concentrations and related processes such as decomposition and nitrogen mineralization. Predicting the types of soil processes and their rates from remote sensing data will depend on strong correlations between canopy variables and the process of interest. This paper discusses the potential for measuring canopy indicators of below-ground processes using current and emerging remote sensing technology, and how we might begin to construct a set of measurements relevant to studies of soil processes and their dynamics over time and space. Basic remote sensing principles are presented with an emphasis on the visible and short-wave infrared spectral region (0.4–2.5 μm), followed by a discussion of the spectral properties of soil and vegetation.

Contextual Pattern Recognition Applied to Cloud Detection and Identification

The use of contextuial information in classification of satellite imagery of clouds is considered. A general contextual classification rule is developed which is then simplified under some realistic assumptions made. for the label relationships of neighboring cloud image pixels. The resulting decision rule is defined in terms of transition probabilities between pairs of neighboring pixels. Experimental evaluation of the rule confirms its superiority over the conventional Bayes rule and provides justification for the assumption made.