Scalable Matching of Agricultural Field Polygons for Large-Scale Data Comparisons

Establishing sustainable agricultural systems while ensuring food security is a global priority. A key requirement for this is detailed crop information, which has become increasingly available through advances in deep learning and data harmonization. Large-scale datasets of agricultural field polygons exist in varying granularity and quality. Identifying correspondences between such datasets covering the same regions opens up important opportunities: for quality assessment through validation of model outputs, for data fusion and integration of complementary information, and for temporal change analysis to monitor developments over time. In this poster, we present our approach for detecting spatial correspondences in large-scale datasets of agricultural field polygons. We show results from two case studies that assess both the spatial and semantic consistency between datasets, thereby demonstrating the suitability and versatility of the algorithm for applications in agricultural science. In addition, we introduce an interactive map viewer that enables an intuitive visual comparison of datasets within their geographic context.