Context and objectives
To evaluate the possibilities of using satellite imagery for agrohydrological studies, in Belgium as well as in developing countries.
Expected scientific results
A complete inventory of plant identification and geographical position of fields of Gembloux has been performed in 1990 and 1991 with the aim of doing image classification tests.
A field's edges automatic detection method based on SPOT panchromatic and multi spectral images has been realised. This method can be used to segment images before doing an automatic per-field classification of the cultivated land. It is also possible to combine images of different standard models and obtained at different dates.
The study of the per-field plant identification by means of discriminant analysis was increased: the variables selection at the level of the monotemporal or multi temporal images data has been examined, and the size of the samples of fields required to compute the discriminant functions were determined.
A post-classification method taking into account the nature of the previous crop canopy as observed in the field was studied, and a strategy was settled. This method was very efficient for plant identification, particularly when satellite data are poor or infrequent.
A simulation of an automatic checking of the crop canopy was realised. That simulation integrates the automatic segmentation, the per-field classification and the post-classification methods. The aim of that trial is to demonstrate the feasibility of an automation of the crop canopy, and to evaluate the precision of the method.
The characteristics of a data processing system architecture able to do image processing in an operational and highly automated mode were determined. Some basic components were developed to satisfy the current needs and to verify the feasibility of such a system.
|Project leader(s):||ULg - Gembloux Agro-Bio Tech - Echanges Eau-Sol-Plante|