An eolian and marine phenomena characterization in dune sectors by image processing

The northern French coastline is partly subject to coastal erosion that results in sandy coast retreat. Many dune sectors are moving because of the wind dynamics due to the small quantity of vegetation on their surface. In order to analyze and control these phenomena that endanger the littoral and tourist activities, several scientific programs have been started that use aerial remote sensing images. The authors study image processing of the coastal environment. The algorithm used recognizes, localizes and quantifies the different components of the littoral environment such as sand features and vegetation. By working with an image set representing northern France and dating from the sixties until today, the algorithm provides the evolution thematic map of the site. From an algorithmic viewpoint, the thematic map is obtained by an unsupervised method of classification which is independent of the ambient luminosity, of the sensor adjustment, of its altitude and of its orientation. This method belongs to the family of SEM algorithms and uses the Bayesian strategy minimizing the classification error. The authors´ algorithms have been applied to a set of images representing the region of Wissant (Pas de Calais, France). As result, they analyze from the thematic map the speed of coastal retreat, the movement of the dunes towards the urban zones and the impact of the plantation program on the coastal environment