Coastline interpretation from multispectral remote sensing images using an association rule algorithm

Abstract

On the basis of the Apriori algorithm, a class association rule algorithm is presented. A sea–land separation method was designed, and then a shoreline detection method proposed for interpreting multispectral remote sensing images. When separating the land from the sea, not only the spectral attributes but also the texture attributes and basic statistical values were considered in attribute space. To test the feasibility of the method, a Landsat Enhanced Thematic Mapper Plus (ETM+) image scene was used to interpret the coastline. First, the association rules of the sea–land separation of the study area were discovered from learning samples by using the class association rule algorithm. Second, the sea and the land of the image were separated with the mined rules. Third, the coastline was interpreted from the separation result. The accuracy of the interpretation result was computed with a proposed line matching accuracy evaluation algorithm. We show that the proposed method can interpret the coastline accurately and does not require any complex preprocessing.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (No. 40906094) and the Project 908-01-WY02 of Marine comprehensive investigation and assessment in China.