ASPRS

PE&RS November 1997

VOLUME 63, NUMBER 11
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
JOURNAL OF THE AMERICAN SOCIETY FOR PHOTOGRAMMETRY AND REMOTE SENSING

Peer-Reviewed Articles

1275 Two-Dimensional Template-Based Encoding for Linear Quadtree Representation
Henry Ker-Chang Chang, Shing-Hua Liu, and Cheng-Kuan Tso

Abstract
Download Full Article
A two-dimensional template-based encoding (2DTE) technique for linear quadtree construction is proposed. The 2DTE technique combines the concept of template mapping and the Morton sequence to encode regional data on an image. With the definition of the last homogeneous pixel, the new coding algorithm can be completed in a time linear to the number of pixels without repetitive scanning of image pixels. Compared with other linear quadtree coding methods, the proposed 2DTE techniqne has a linear n time reduction of starage space if a 2^n by 2^n image is processed. The potential of the proposed 2DTE technique for encoding multicolored images is also described. Several empirical tests and theoretical analyses verify that the proposed 2DTE technique outperforms other linear quadtree construction algorithms. The proposed 2DTE technique is believed to be profitable for applications in image processing or geographic information systems.

1285 An Evaluation of the Potential for Fuzzy Classification of Multispectral Data Using Artificial Neural Networks
Timothy A. Warner and Michael Shank

Abstract
Download Full Article
Fuzzy classification, or pixel unmixing, is the estimation of the proportion of the cover types from the composite spectrum of a mixed pixel. In this paper, we evaluate how the separation between class means, the covariance matrix of each class, and the relative location of the class means in the spectral space limit the fuzzy representation of mixtures. The influence of these factors is ilInstrated with a fuzzy classification using a back-propagation artificial neural net. Experiments using simulated data indicate that a fuzzy classification with an average error of less than 10 percent requires a Bhattacharyya Distance between classes of at least 9. The error in the fuzzy represen tation using a neural net also varies as the proportions of the classes changes. with a peak error when one class comprises approximately 0.20 to 0.25 of the mixed pixel. Back-propagation neural networks are not necessarily good at spectral unmixing. The backpropagation neural network produces spectral-space partitions between the classes that are generally steep, and that are not necessarily midway between the classes. The partitions tend to be simple, and somewhat linear. In addition, the output on nodes does not have to sum to 1.0, which may result in situations where high values are predicted for two classes simultaneously. Two methods of improving neural network behavior for fuzzy classification include use of a compound linear-sigmoid activation function, and mining using synthetic mixed pixels. 

1295 Automated Vegetation Mapping Using Digital Orthophotography
Roland J. Duhaime, Peter V. August, and William R. Wright

Abstract
Download Full Article
We used near-infrared digital orthophotography and three collateral data sets to model ecological communities on Block Island, Rhode Island. Aerial photograpy of the island was taken on 19 May 1992 at a scale of 1:40,000. The photography was scanned and processed to remove distortions from terrain, aircraft , and optical aberration. The resulting digital orthophotograph was comprised of three spectral bands representing the red, green, and blue colors of the scanned photography and had a pixel dimension of 1.27 m. Three textural variables were developed by calculating the standard deviation within a 10-m radius of every pixel for each of the three spectral bands in the image. The terrain model that was used to create the orthophoto was also used to derive slope and aspect for each pixel. Soil survey data were used to map the distribution of soil drainage classes to distinguish wetland from upland vegetation. We used linear discriminant analysis to develop a model to distinguish 11 vegetation and cover classes on the island. The fill model consisted of nine independent variables derived from the orthophoto, the textural indices, terrain metrics, and soils. Classification accuracies ranged from 60 to 80 percent for an independent validation data set. The variable DRAINAGE CLASS dominated the model and explained the most variation in vegetation and cover class.

1303 A Comparison of Nighttime Satellite Imagery and Population Density for the Continental United States
Paul Sutton, Dar Roberts, Chris Elvidge, and Henk Meij

Abstract
Download Full Article
The striking apparent correlation between nighttime satellite imagery and human population density was explored for the continental United States. The nighttime stable-lights imagery was derived from the visible near-IR band of 231 orbits of the Defense Metrological Satellite Program Operational Linescan System (DMSP-OLS). The population density data were generated from a gridded vector dataset of the 1992 United States census block group polygons. Both datasets are at a one-square-kilometer resolution. The two images were co-registered and correlation between them was measured at a range of spatial scales, including aggregation to state and county levels. DMSP imagery showed strong correlations at aggregate scales, and analysis of the saturated areas of the images showed strong correlations between the areas of saturated clusters and the populations those areas cover. The non-zero pixels of the DMSP imagery correspond to only 10 percent of the land cover yet account for over 80 percent of the cantinental United States population. Spatial analysis of the clusters of the saturated pixels predicts population with an R² of 0.63, Consequently, the DMSP imagery may prove to be useful to inform a "smart interpolation " program to improve maps and datasets of human population distributions in areas of the world where good census data may not be available or do not exist.
Top Home