Dynamic High-Precision Field Shape Generation via Combine GPS Tracks

Abstract. Fields are in practice the most fundamental elements for agricultural management. Crops within the same field are usually not only treated similarly as a group during planting, growing and harvesting, they are also often analyzed together to represent the performance of that particular field. Currently, fields are often defined as areas inside some static field boundaries previously decided by the operator. However, the shape for the actual area in a field which yields valid products may change dramatically over time because of weather, natural terrain changes, in-field constructions and field management decisions. And the current static approach to defining fields makes it hard for comprehensive high-precision performance evaluation. In this paper, we propose to utilize GPS tracks for combine harvesters during the harvesting season to extract up-to-date, high-precision field shapes. We use the notion of α-shape to outline the fields and detect holes inside them. After that, another technique, statistical replay for harvesting, is used to reasonably extend the field shapes. The whole algorithm is fully automatic and the field boundaries generated by it agree with the ones provided by the field owners, but capture far more details, like holes, which are parts not suitable for farming, inside the fields. We also looked at some possible application cases for our algorithm. Results for the same fields in different years are compared, and the differences we see in fact reflect several important decisions made by the farmers.