In modern agriculture, machinery has become the one of the necessities in providing safe, effective and economical farming operations and logistics. In a typical farming operation, different machines perform different tasks, and sometimes are used together for collaborative work. In such cases, different machines are associated with representative activity patterns, for example, in a harvest scenario, combines move through a field following regular swaths while grain carts follow irregular paths as they ferry grain from combines to trucks. Sometimes unusual conditions due to field conditions, machine status, weather, or human factors may cause anomalous activity patterns. Detecting and classifying anomalous patterns can be used for planning and efficiency improvements.

Zhang performed rudimentary work (Zhang et al. 2017) utilizing GPS paths for classifying different machines’ activity patterns using a rule-based algorithm to understand what has happened in the field for farm logistics improvements. In this paper, we use machine data in addition to GPS tracks for anomaly detection and classification. The algorithm uses Kalman filtering and the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm (Ester et al. 1996) to cluster a point cloud that consists of a combine’s engine load, speed data, and the Kalman filter residual.