An internship at the VITO in Belgium. Estimating agricultural area in Senegal from medium resolution images, using an up-scaling method based on Artificial Neural Networks.
I conducted my thesis at VITO, which stands for “Vlaams Instituut voor Technologisch Onderzoek’ or in English: ‘Flemish Institute for Technological Research’, within the framework of the Global Monitoring for Food Security (GMFS) project (www.gmfs.info), which provides, amongst others, early warning, yield assessments and agricultural mapping services for food security monitoring in sub-Saharan Africa.
To avoid famines, crop production should be known as early as possible in the growing season. This enables local authorities and international institutes to assess the scale of the problem, examine alternatives and decide on actions.
Crop production is basically predicted by multiplying yield estimates with crop area estimates, the latter being the aim of my thesis. For crop area estimations in vast agricultural regions, satellite images are indispensable. Medium resolution images are preferred to high resolution images, since they cover a larger area and are less expensive. However, the field size is mostly smaller than the pixel size, resulting in mixed pixels. Therefore sub-pixel classification methods have to be used, such as ‘Artificial Neural Networks’ (ANN), which I used.
GMFS is active in Ethiopia, Sudan, Senegal, Zimbabwe and Malawi. We preferred Senegal as research area for my thesis, since it has the largest and most regular fields, which facilitates the recognition on satellite images. If this ‘pilote study’ yields good results, the techniques of this study can be further refined working with satellite images of the other countries.
I started my thesis in September with the collection of ground truth data. Since it was the aim to get as much representative data for the whole of Senegal as possible, I traversed large parts of the country by car, together with a local expert. The followed tracks were logged with a GPS. The crops types we crossed along the road were registered together with the corresponding coordinate.
Once the field data were collected, I could start my research at VITO. I designed different ANNs . I was surrounded by experts working on the same project or working with the same techniques.
Being back in Belgium, the working experience at VITO was very interesting. They provided me with all the facilities and expertise to finally apply the knowledge about remote sensing I gathered during the master courses. After my research was finished, I had the chance to present my findings on the user board meeting.
At last, with the MGI thesis I acquired a lot of practical skills and had a unique experience abroad as well as in a renowned research centre.