Watershed regions and watershed lines based cooperation strategy for image segmentation. Application to roof detection

In this paper, we present a strategy of image segmentation for roof detection from aerial images. For that, an orthophotoplan segmentation method based on a cooperation technique including edge based segmentation and region based segmentation is proposed. Both segmentation techniques are assured by watershed algorithm. The proposed strategy is composed of three steps: (i) A simplification step that consists in simplifying the image with an appropriate couple of invariant/gradient optimized for the application, (ii) The second step is composed of two main treatments: On the one hand, the image is segmented by the watershed regions coupled with an efficient regions merging technique. On the other hand, the image is segmented by the watershed lines, (iii) The third step consists in exploiting the advantages of each method and thus achieve a reliable segmentation results. Tests have been performed on an orthophotoplan containing 100 roofs with varying complexity and evaluated with Vinet criteria using a ground truth image.