The importance of automatic classification procedures applied on digital images (aerial orthophotographs, satellite images, etc.) for the recognition of soil coverage or morphological elements (geological discontinuities, ground textures, etc.) is widely confirmed by several studies on environmental issues. As far as crop shelters are concerned, there is no thematic cartography generally available mainly due to the temporary character of this type of constructions. Yet a wide and deep knowledge of localization and territorial density of crop shelters can be of great interest in strategic choices for environmental and landscape planning and protection. Aim of this study is to verify the effectiveness of image processing application to the classification of crop shelter sites and its accuracy. An intensively-cultivated area located in South-Eastern Sicily was selected to perform the localization of greenhouses and other type of shelters (e.g. vineyards coverings). A mosaicked digital orthophotograph of Sicilian territory was subset to obtain the study area. This image was georeferenced and analyzed by an image processing software. A methodology based on supervised classification of the image was found as the most adequate to the classification of crop shelters. According to this methodology suitable classes were selected on the basis of signatures related to specific sample areas. The classification was then refined by using neighbourhood and contiguity analysis algorithms. The results of the analyses allowed to recognize, localize the crop shelters and quantify their planimetric area. The latter was also compared with the attributes of georeferenced feature classes based on visual recognition. The application of this methodology to adjacent areas showed good results. The methodology developed in this study could provide the technical cartography with a new thematic layer of crop shelter coverage. This application, among the others, could be useful to show the modifications in crop shelter localization during years in a specific area.
