KITE
The aim of this research project is recognizing Kites on satellite images using a graph based approach. Kites are remnants of long stone walls that outline the shape of a child’s kite. But the kites are huge, their big size makes them often clearly visible on high-resolution satellite images. A large scale recognition of kites on their wide distribution area, will help archeologists of various fields to understand these enigmatic constructions. The Kite project makes progress on the basis of cooperation with the ANR project Global Kite. An important aspect of the project is the development of a pattern recognition tool that will automatically find kites on satellite images. This tool will allow a comprehensive inventory of these structures worldwide. From this inventory, a study is conducted, from the Arabian peninsula to the Caucasus, on the relationship between habitat areas and areas of kites combined with an identification of all the zooarchaeological data available in the literature to analyze the issue of kites on the economies subsistence and the importance of hunting and farming in the early urban societies. The systematic identification of kites on satellite images enables to delimit spatially this phenomenon. It also identifies other hunting or pastoral developments that are analogous to kites but must be culturally excluded. Large scale kite identification can only be carried out by a gradual work in close collaboration with archaeologists, geographers and computer scientists. A first model for kite recognition will be established based on the archaeological and geographical knowledge. This model will be enriched progressively with the feedback of archaeologists and geographers experts on the results provided by the model. Object recognition in images is an area of intensive research worldwide, and especially in recent years. This research has many locks, some of which may be considered in this project, such as technical fast graph matching, modeling images by graphs and especially taking into account the information of color, shape and texture in the graph models used for kite recognition in satellite images.