Cartographie forestière à l'aide d'images radar et optiques multirésolution cas de la province de Phu Tho, Vietnam
Luong, Anh Tuan
This study plans to produce a map of the forest cover of the central part of Ha Hoa district, Vinh Phu Province, northern Vietnam, by using a spaceborne multi sensor approach. It has two parts: The first part presents a bibliographic search in the optical and radar domains. Remote sensing is still a recent technology in Vietnam, especially in the Radar domain. Definitions and characteristics of Remote sensing are presented in order to show the relationships between Remote sensing technologies and image processing in order to classify landscape and terrain features such as forests, agriculture, soils and water. We have also presented the data fusion theory in order to select the context based fusion technique to integrate the classification results of two images. The second part presents an application of the classification and fusion technique for LANDSAT TM and RADARSAT images. The results are validated through analysis of field data. Non-supervised classification is used for LANDSAT TM data and a supervised approach of textural channels is used for the RADARSAT image. The contextual fusion of classification results is based on the response of each sensor for the specific targets of the study area. Validation of results with terrain data shows the utility of the different image types for mapping the extent of forest cover. This validation shows also that the fused data set gives better results than the individual LANDSAT and RADARSAT images.