期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2007
卷号：XXXVI-1/W51
出版社：Copernicus Publications
摘要：As a contribution to a better understanding of the integration of different data sets into knowledge-based classification processes the presented study compares different data-fusion techniques which combine spectral and textural information of a QuickBird scene with ancillary data layers and a knowledge base for the identification of forest structures and habitats. The approach compares a fuzzy logic classifier and a crisp rule-based integration technique. Both methods combine spectral and textural information with ancillary data-layers and a knowledge base. For the fuzzy logic approach, the possibility of assignment for each object is combined with a fuzzy knowledge base, which consists of information about the possibility of existence of tree species and geo-factors for a GIS-database consisting of slope, aspect, and height of a medium resolution DTM, soil type, and forestry site maps. Therefore, for all woodland species of this specific region, an index of location factors was developed. A fuzzy set for each class concerning each geo-factor is defined, containing membership functions. In order to develop a better understanding of the integration patterns of single geo-factors a significance analysis was carried out. This was done by an ISODATA Clustering and a significance analysis of microarrays (SAM), which is abundantly applied in Bioinformatics. A sequence of classifications with various applied geo-factors for different classes was performed. Accuracy assessments based on test samples were calculated and the results arranged in a microarray. The results from the ISODATA Clustering and the SAM showed a high significance of the combined utilization of all additional geo-factors. Moreover, classes with smaller percentages of covered area and smaller ecological niches depend more on the application of ancillary information than other forest types. Additionally, two types of geo-factors were detected as insignificant, which is due to an inappropriate data quality of a soil-map and the lack of significance of the geo-factor aspect
