High Performance Real-time Anomaly Detection for Agricultural Monitoring Systems

Due to technological advancements, real-time monitoring of various agricultural parameters has become essential to modern farming practices. Monitoring agricultural parameters such as temperature, humidity, soil moisture, and plant growth helps farmers to optimize their farming practices to improve crop yields. However, near-real-time monitoring involves significant amounts of data, and analyzing this data in real time for anomalies is a challenging problem. In this paper, we propose a novel approach for anomaly detection in near real-time agriculture monitoring using a cortical coding method to address this issue. The cortical coding method utilizes a tree structure to cluster and learn patterns from time-series data. The proposed method has previously proven helpful in tasks such as vector quantization and anomaly detection. This paper explores the application of cortical coding-based anomaly detection in agriculture monitoring. Incremental and continuous learning is highly beneficial for early and accurate detection in systems with large data streams. The proposed method is evaluated on a dataset obtained from an agricultural monitoring station in Turkey. Experiments show that the proposed method outperforms widely used forecast-based anomaly detection methods.