Mapping Coastal Vegetation Using an Expert System and Hyperspectral Imagery

Abstract

Mapping and monitoring salt marshes in the Netherlands are important activities of the Ministry of Public Works (Rijkswaterstaat). The Survey Department (Meetkundige Dienst) produces vegetation maps using aerial photographs. However, it is a time-consuming and expensive activity. The accuracy of the conventional vegetation map derived using aerial photograph interpretation (API) is estimated to be approximately 43%. An alternative method is demonstrated that uses an expert system to combine airborne hyperspectral imagery with terrain data derived from radar altimetry. The accuracy of the vegetation map generated by the expert system increased to 66%. When hyperspectral imagery alone was used to classify coastal wetlands, an accuracy of 40% was achieved - comparable to the accuracy of the API-derived vegetation map. An analysis of the efficiency of the proposed expert system showed that the speed of map production is increased by using the new method. This means that digital image classification using the expert system is an objective and repeatable method superior to the conventional API method.