應用於衛星影像之自動化土地利用分析系統

Abstract

Because satellite images are easy to access at present, their use for various applications has become an interesting and important research issue. The objective of this paper is to develop an automatic land use analysis system using satellite images. The methods include pre-processing, similar area labeling, and land use analysis. The core technique is the (α, ω)-connected component labeling that is used to classify the land in images into “developed area” and “undeveloped area”, respectively. Results are evaluated using manual classification by a geologist and the automatic classification by the system. Our system is shown to achieve an accuracy of ~90%, which clearly demonstrates that our system could achieve comparable results by geo-technologists. In addition, our system can process the satellite image in ~10 seconds in comparison to 15~30 minutes required by geologists, thus may ultimately be used as a computer-aided system to simplify the labor-intensive tasks by human for the land use analysis.