ASPRS

PE&RS August 2005

VOLUME 71, NUMBER 8
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
JOURNAL OF THE AMERICAN SOCIETY FOR PHOTOGRAMMETRY AND REMOTE SENSING

Peer-Reviewed Articles

909 Bias-compensated RPCs for Sensor Orientation of High-resolution Satellite Imagery
Clive S. Fraser and Harry B. Hanley

Abstract  Download Full Article
The demand for higher quality metric products from high-resolution satellite imagery (HRSI) is growing, and the number of HRSI sensors and product options is increasing. There is a greater need to fully understand the potential and indeed shortcomings of alternative photogrammetric sensor orientation models for HRSI. To date, rational functions have proven to be a viable alternative model for geo-positioning, and with the recent innovation of bias-compensated RPC bundle adjustment, it has been demonstrated that sensor orientation to sub-pixel level can be achieved with minimal ground control. Questions have lingered, however, as to the general suitability of bias-compensated rational polynomial coefficients (RPCs), and indeed rational functions in general. The purpose of this paper is to demonstrate the wide applicability of bias-compensated RPCs for high-accuracy geopositioning from stereo HRSI. The case of stereo imagery over mountainous terrain will be specifically addressed, and results of experimental testing of both Ikonos and QuickBird imagery will be presented.

917 A Dynamic Method for Generating Multi-Resolution TIN Models
Bisheng Yang, Wenzhong Shi, and Qingquan Li

Abstract  Download Full Article
It is essential to generate multi-resolution Triangulated Irregular Network (TIN) models dynamically and efficiently in three-dimensional (3D) visualization, virtual reality, and geographic information systems (GIS), because the data that needs to be processed is multiple in scale and large in volume. This paper proposes a new method, which extends the edge collapse and vertex split algorithms, to dynamically generate a multi-resolution TIN models. In contrast to previous approaches, a new method is proposed to encode and store vertex dependency relationships in the multi-resolution model. As a result, the validity of vertex splits and edge collapses is improved; the efficiency of storing data is also enhanced by the proposed method. To evaluate the performance of the proposed method, we further extend the assessment to (a) time cost; (b) the quality of the multi-resolution TIN model; and (c) the view-dependent multi-resolution model. The root mean square error (RMSE) of the elevation of the vertex and the quality of the shape of the triangle are adopted to evaluate the quality of a generated multi-resolution TIN model. The results of the experiment demonstrate that the proposed method performs better than previous methods in terms of time cost, and can achieve multi-resolution TIN models with a higher accuracy.

927 Examining Lacunarity Approaches in Comparison with Fractal and Spatial Autocorrelation Techniques for Urban Mapping
Soe W. Myint and Nina Lam

Abstract  Download Full Article
The conventional spectral-based classification techniques have often been criticized due to the lack of consideration of images’ spatial properties. This study evaluates and compares two lacunarity methods, fractal triangular prism, spatial autocorrelation, and original spectral band approaches in classifying urban images. Results from this study show that the traditional spectral-based classification approach is inappropriate in classifying urban categories from high-resolution data. The fractal triangular prism approach was also found to be ineffective in classifying urban features. Spatial autocorrelation was more accurate than the fractal approach. The overall accuracies in this study for the fractal, conventional spectral, spatial autocorrelation, lacunarity binary, and lacunarity gray-scale approaches were 52 percent, 55 percent, 78 percent, 81 percent, and 92 percent, respectively. These findings suggest that the lacunarity approaches are far more effective than the other approaches tested and can be used to drastically improve urban classification accuracy.

939 Fuzzy Reliability Assessment of Multi-Period Land-cover Change Maps
Kim Lowell, Gary Richards, Peter Woodgate, Simon Jones, and Laurie Buxton

Abstract  Download Full Article
A fuzzy methodology is presented for evaluating the reliability of satellite imagery-derived, continent-wide (Australia) land-cover deforestation/regrowth maps covering the period of 1972 to 2000 in ten discrete time periods. The methodology uses aerial photographs as its reference data and accommodates the difficulty inherent in determining definitively from an aerial photograph, whether a sample point is Forest or Non-forest by permitting interpreters to identify their level of certainty, i.e., Definitely Forest, Probably Forest, Uncertain, Probably Non-forest, or Definitely Non-forest. This information is then cross-tabulated against the Forest/Non-forest classification for the classified image closest in date to the photo date. Information from several photographs is summarized over a larger geographic area and over all time periods. Subsequently, temporal lineage information for each sample pixel is extracted from the 1972 to 2000 series of classified images to determine if a pixel’s lineage is Forest Throughout, Non-forest Throughout, Deforestation, Regrowth, or Cyclic. The fuzzy evaluation for individual pixels is then tabulated against this lineage information to identify if pixels of any particular lineage have an elevated tendency to be misclassified. The methodology provides a means by which problems in the map production methodology can be improved as future time slices are added.

947 Using Landsat ETM+ Imagery to Measure Population Density in Indianapolis, Indiana, USA
Guiying Li and Qihao Weng

Abstract  Download Full Article
Remote sensing techniques have been previously used in urban analysis, settlement detection, and population estimation. This research explores the potentials of integration of Landsat ETM+ data with census data for estimation of population density in City of Indianapolis, Indiana. Spectral signatures, principal components, vegetation indices, fraction images, textures, and temperature were used as predictive indicators. Correlation analysis was used to explore the relationships between remote sensing variables and population, and stepwise regression analysis was then used to develop models for estimating population quantities. Two sampling schemes (non-stratified versus stratified sampling) were compared. It was found that the integration of textures, temperatures, and spectral responses substantially improved the accuracy of estimation. Stratification of the population into three categories of low-, medium-, and high-densities and development of different models for individual population density category provided better estimation results than a non-stratified scheme. The total population for City of Indianapolis was estimated to be 832,792 in 2000 yielding an accuracy of 96.8 percent.

959Characteristics of Seasonal Vegetation Cover in the Conterminous USA
Kevin Gallo, Brad Reed, Timothy Owen, and Jimmy Adegoke

Abstract  Download Full Article
A data set of the fractional green vegetation cover (FGREEN) for the Conterminous USA was evaluated for regional and seasonal variation. The value of FGREEN was derived monthly for the three most dominant land cover classes per 20 km by 20 km grid cell within the study area. At this grid cell resolution (comprised of 400 1-km pixels), 97 percent of the grid cells included three or fewer land cover classes. FGREEN was found to vary regionally due to local land cover and climate variations. FGREEN was found significantly different between one or more of the land cover classes, for one or more months, in 58 percent of the grid cells included in the study. Monthly FGREEN values for the land cover classes vary sufficiently between the land cover classes to warrant monthly FGREEN data for each of the one to three most dominant land cover classes per grid cell.

967 Satellite Estimation of Aboveground Biomass and Impacts of Forest Stand Structure
Dengsheng Lu, Mateus Batistella, and Emilio Moran

Abstract  Download Full Article
Heterogeneous Amazonian landscapes and complex forest stand structure often make aboveground biomass (AGB) estimation difficult. In this study, spectral mixture analysis was used to convert a Landsat Thematic Mapper (TM) image into green vegetation, shade, and soil fraction images. Entropy was used to analyze the complexity of forest stand structure and to examine impacts of different stand structures on TM reflectance data. The relationships between AGB and fraction images or TM spectral signatures were investigated based on successional and primary forests, respectively, and AGB estimation models were developed for both types of forests. Our findings indicate that the AGB estimation models using fraction images perform better for successional forest biomass estimation than using TM spectral signatures. However, both models based on TM spectral signatures and fractions provided poor performance for primary forest biomass estimation. The complex stand structure and associated canopy shadow greatly reduced relationships between AGB and TM reflectance or fraction images.

975 Comparing Raster Map Comparison Algorithms for Spatial Modeling and Analysis
Matthias Kuhnert, Alexey Voinov, and Ralf Seppelt

Abstract  Download Full Article
The comparison of spatial patterns is recognized as an important task in landscape ecology especially when spatially explicit simulation modeling or remote sensing is applied. Yet, there is no agreed procedure for doing that, probably because different problems require different algorithms. We explored a variety of existing algorithms and modified some of them to compare grid-based maps with categorical attributes. A new algorithm based on the “expanding window” approach was developed and compared to other known algorithms. The goal was to offer simple and flexible procedures for comparing spatial patterns in grid based maps that do not take into consideration object shapes and sizes of the maps. The difference between maps was characterized by three values: quantity, location, and distance between corresponding categories in the maps. Combinations of these indices work as good criteria to quantify differences between maps. A web-based survey was set up, in which participants were asked to grade the similarity of ten pairs of maps. These results were then used to compare how well the various algorithms can perform relative to the visual comparisons obtained; they were also used to calibrate existing algorithms.

Top Home