A synergistic approach comprising high resolution multispectral UAV optical dataset and synthetic aperture radar (SAR) can help in understanding biophysical details of target objects especially with crop type identification. Ensuring food security in Africa requires accurate crop type mapping of agricultural regions. The International Institute of Tropical Agriculture, Ibadan (IITA, CGIAR) is at the forefront of research for sustainable agriculture and food security in Africa. This work conducted on the agricultural parcels of the institute availed high spatial resolution (12cm, resampled to 50cm) multispectral UAV data in combination with the Sentinel 1C Synthetic Aperture Radar (SAR) dataset. Eleven vegetation indices have been generated from the multispectral image of the UAV post which a principal component analysis was carried out to reduce the feature space. A Support Vector Machine (SVM) and Random Forest (RF) classifiers were used to model and classify the different crop types. A combination of VV and VH polarizations of Sentinel SAR data are also analyzed to classify the crop types while comparing and contrasting its accuracy with the UAV data derived products. From the analysis, the SVM produce an overall classification accuracy of 94.78% and 81.72% for UAV and SAR datasets respectively while the RF produce an accuracy of 93.84% and 92.58% respectively for UAV and SAR datasets. Furthermore, simple models’ agreement ratio is computed which also produces agreement ratio above 90% in some cases. The UAV data modelling provides an improved accuracy when compared to similar object-based classification on UAV data. The models agreement presented would also be useful as a quick tool to assess models’ agreement in a multi-modelling scenario. The combined approaches are useful in precision agriculture over small and large agricultural fields to ensure food security.
Keywords: Support vector machine, random forest, precision agriculture, UAV, SAR, IITA, Nigeria.