Detection of Wooded Hedgerows in High Resolution Satellite Images using an Object-Oriented Method

Abstract

The objective of this study was to identify wooded hedgerows from remote sensing data in using an object-oriented approach, in order to estimate the proportion of hedgerow network that can be automatically extracted, whatever its characteristics. To evaluate the reliability, accuracy, and computational efficiency of the object-oriented method to extract wooded hedgerows, we applied it on different types of remote sensing images on six study sites located in bocage landscapes of Northern-western France. These images were segmented on three hierarchical levels (tree, hedge and field) and were subsequently classified by means of membership functions using fuzzy logic. The results show that the remote sensing images with a spatial resolution equal or less than 10 meters are appropriate to automatically inventory wooded hedgerows. The results also highlight that agricultural landscape complexity influences the classification accuracy, as the detection performance increases with hedges density.