Significance Analysis of Different Types of Ancillary Geodata Utilized in a Multisource Classification Process for Forest Identification in Germany

Ancillary geodata can supply information to enhance classification accuracy for a variety of remote-sensing applications. To understand the integration of different data into a knowledge-based multisource classification process, this paper evaluates the significance of geodata for the classification accuracy of a very high spatial resolution satellite image for the identification of forest types in Germany. The approach utilizes a fuzzy-logic classifier for the integration of a knowledge base, which combines spectral information with ancillary data layers. The results of the classification were used to test a method for evaluating the influence of the integration of single geodata, the effects on different classes, and the impacts of the applied rules. A microarray significance analysis (MSA) was used to evaluate the significance of the classification results, whereas an ISODATA clustering was utilized for visualizing. A sequence of 50 accuracy assessments of classifications with possible combinations of geodata and rules for the identified classes was derived. The resulting microarray of accuracy percentages of single classes and the overall classification was used for further investigation. The MSA supplies the measure of significance, called relative difference d(i). The MSA identified 11 classifications of positive significance (d(i) greater than 1.44) and three classifications of negative significance (d(i) lower than -2.87). In particular, classifications that contain all rules were rated as positive significant.