ASPRS

PE&RS August 1998

VOLUME 64, NUMBER 8
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
JOURNAL OF THE AMERICAN SOCIETY FOR PHOTOGRAMMETRY AND REMOTE SENSING

Peer Reviewed Articles

797 Window Size for Image Classification? A Cognitive Perspective
Michael E. Hodgson

Abstract
Download Full Article
Windows are commonly used in digital image classification studies to define the local information content around a single pixel using a per-pixel classifier. Other studies use windows for characterizing the information content of a region, or group of pixels, in an area-based classifier. Research on identifying window size and shape properties, such as minimum, maximum, or optimum size of a window, is almost exclusively based on the results from automated classifications. Under the notably different hypothesis about optimum sizes of windows in automated classifications and approaches for determining such optimum window size, this article presents a cognitive approach for evaluating the functional relationship between window size and classification accuracy. Using human subjects, a randomized experimental design, and a continuum of window sizes, portions of digital aerial photographs were classified into urban land-use classes. Unlike the findings from purely automated approaches, classification accuracy from visual analysis increased in a monotonic form with increasing window size for the urban land-use classes investigated. A minimum window size of 40 by 40 pixels (60-m by 60-m ground area) was required for classifying Level II urban land use using 1.5-m by 1.5-m resolution data (>=75 percent accuracy. This finding is dramatically different from the ideal window size range (i.e., 3 by 3 to 9 by 9) and functional relation between window size and classification accuracy found in automated per-pixel classifications. A theoretical curve depicting the relationship between classification accuracy and window size, spatial resolution, end classification specificity is presented. 

809 Teaching the Physical Principles of Vegetation Canopy Reflectance Using the SAIL Model
F.M. Danson

Abstract
Download Full Article
A wide range of resources are available to remote sensing teachers to introduce students to the principles and applications of image processing, but, in contrast, there are few resources suitable for teaching the physical principles of the subject. This paper describes how a radiative transfer model of vegetation canopy reflectance may be used to allow students to explore the complex set of factors that control vegetation canopy reflectance. An example of a practical exercise used with undergraduate level students is described, and topics for follow-up discussions are outlined. The model may be obtained from http://www.salford.ac.uk/geog/staff/sail.html and used without restriction. 

813 Investigation of the Integration of AVIRIS and IFSAR for Urban Analysis
George F. Hepner, Bijan Houshmand, Igor Kulikov, and Nevin Bryant

Abstract
Download Full Article
Most attempts at urban analysis using remotely sensed imagery lack the capabilities necessary to define the detailed geometry and differentiate the textures of the complex urban landscape. This paper presents a proof-of-concept study of the potential for integrative analysis of Interferometric Synthetic Aperture Radar (IFSAR) and Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) hyperspectral imagery for a study area in Los Angeles, California. Recent advances in the use of interferometric radar allow the definition of high resolution, three-dimensional (3D) geometry of surface features, topography, end impervious surfaces in urban areas. The radar analysis is enhanced using hyperspectral imagery to mask surfaces adjacent to structures in order to assist in the determination of baseline topography and segmentation of building footprints for improved geometric measurement of the complex urban area. 

821 North American Landscape Characterization Dataset Development and Data Fusion Issues
Ross S. Lunetta, John G. Lyon, Bert Guindon, and Christopher D. Elvidge

Abstract
Download Full Article
With the launch of Landsat 1 on 23 JuIy 1972, the United States initiated the capability for land resource monitoring from space. Over a 20-year period, the Landsat Multi-Spectral Scanner (MSS) sensor collected data documenting landcover conditions over the majority of the globe. The global change research community has prioritized reducing the uncertainty associated with land cover-change as a major constituent of importance to balancing the global carbon cycle. The MSS archive currently represents the best available,continuons public source of relatively high resolution imagery for the monitoring of land cover over the 1972-1992 period. The North American Landscape Characterization project was designed to exploit this archive by providing standardized satellite data sets to support land-cover change analysis. Land-cover categorization products derived from MSS data alone provide only a general characterization of land-cover condition and change. A data fusion approach using a postclassification technique is presented as a cost-effective method for delineating land cover at appropriate thematic and spatial resolutions to support a quantitative inventory of forest carbon stocks. Issues related to assessing the accuracy of carbon inventory datasets is presented, and a two-step model is proposed for accuracy assessment. 

831 Map-Guided Classification of Regional Land Cover with Multi-Temporal AVHRR Data
David M. Stoms, Michael J. Bueno, Frank W. Davis, Kelly M. Cassidy, Kenneth L. Driese, and James S. Kagan

Abstract
Download Full Article
Cartographers often need to use information in existing landcover maps when compiling regional or global maps, but there are no standardized techniques for using such data effectively. An iterative, "map-guided" classification approach was developed to compile a spatially and thematically consistent, seamless land-cover map of the entire Intermountain Semi-Desert ecoregion from a set of semi-independent subregional maps derived by various methods. A multi-temporal dataset derived from AVHRR data was classified using the subregional maps as training data. The resulting regional map attempted to meet the guidelines of the proposed National Vegetation Classification System for classification at the alliance level. The approach generally improved the spatial properties of the regional mapping, while maintaining the thematic detail of the source maps. The methods described may be useful in many situations where mapped information exists but is incomplete, has been compiled by different methods, or is based on inconsistent classification systems.
Top Home