Quality of Building Extraction from IKONOS Imagery

Abstract

Now to ensure the quality of automatic extraction of buildings from aerial and space imagery is not comprehensively understood. Dedicated efforts are needed in examining both the theoretical and practical aspects of the problem. This paper first presents a review of building extraction research activities, with the focus on quality-related issues. Also addressed is recent progress in the use of high-resolution satellite images. As a theoretical study, a set of five quality measures of the extracted buildings is proposed that considers all possible types of errors and describes the accuracy of building extraction relative to the ground truth. To achieve a reliable and precision result, a step-by-step quality control strategy is implemented in a classification-guided building extraction procedure. Unreliable results are identified based on the proposed quality index and excluded from further processing. Empirical results using IKONOS images are presented to demonstrate the proposed quality measures and quality control strategy.