Apport de la télédétection spatiale et des systèmes d'information géographique à la préparation d'un cadastre polyvalent cas de la région de Niamey (Niger)
The installation of a land register with multiple goals, requires the establishment of a basic chart with average scale (between 1:50 000 and 1:25 000), representative in a total and more or less precise way the territory. The techniques generally used in the developing countries are slow and expensive to achieve this task (raised land surveying, photogrammetry and photo-interpretation); what thus produces constraints of duration of implementation, of cost of realization and update of this basic chart. Vis-a-vis to all these constraints, the remote sensing and the geographical information systems bring new solutions at relatively low cost, in the extraction of information on the natural resources and the environment starting from aerospace images, as well as the analysis, the interpretation and the management of this information. The methodological step consists of the combination of geographical information and a technique of segmentation set of themes by spectral-texturale co-operation of images HRV of SPOT, within the framework of a Geographical Information system GIS. We carried out a classification directed to the maximum of probability on the spectral channels panchromatic and, combined with three channels of texture and process to the digitalization of the topographic card and the card of contours of the fields raised with the GPS of the site of study; these data are then integrated in the GIS in the process of establishment of the basic chart of the area of study. The results show, for classification: a total precision of 87% for the classes taken into account. For the evaluation of the surfaces, we notice statistics of agreement between the surfaces measured on the basic image and those measured on the ground being located between 85% and 99%. We obtain from this fact an average precision of 95% and a difference of average surface of 1,3 ha for the whole of the samples of fields. According to the scales, the desired precision is reached in 80% of the samples of fields on scale 1:25 000, and 93% on scale 1:50 000.