Multi-source Information Fusion: Monitoring Sugarcane Harvest Using Multi-temporal Images, Crop Growth Modelling, and Expert Knowledge

This paper deals with the automatic detection of sugarcane harvesting using multi-source information fusion. Information extracted from multi-temporal imagery is fused with indicators from crop growth modelling, and are combined with expert knowledge. The introduced decision support system uses the fuzzy sets theory to cope with uncertainty and imprecision. Fuzzy inference is based on Mamdani's method. The output belongs to three possible classes, and it is accompanied by membership values. The system was evaluated on an irregular time series of SPOT5 images acquired on Reunion Island with significant acquisition gaps. Daily climatic data were used to run the growth model. Results obtained were satisfactory; an overall accuracy of 93% is obtained.