Fertilization zones generation using satellite imagery

Abstract

One of the main goals of precision agriculture is to maximize crop yields, which requires efficient utilization of fertilizers. We present an automated procedure for generating a vector layer of zones that require different intensity of fertilization based on the value of vegetation index obtained from Sentinel-2 imagery. After obtaining the images we create a histogram, where the number of bins determines the number of fertilization zone classes we wish to obtain. Using the marching squares algorithm, we produce a final vector layer, which is exported to the GeoPackage format, making it compatible with modern agricultural equipment. We observe that fertilization zone generation depends on the field size, as large fields can generate more compact zones. Smaller fields tend to result in fragmented zones, due to bigger influence of neighboring regions pixels intensities. This approach thus offers a scalable solution for generation of fertilization zones, that can be used in fertilization map generation, but adjustments would be needed to reduce the effect of fragmentation.