A two-stage agriculture environmental anomaly detection method

Abstract

In order to process abnormal problems of massive distributed greenhouse environmental data, a novel anomaly detection algorithm based on the combination of Support Vector Machine (SVM) and Gaussian Mixture Model (GMM) is proposed and realized under the Spark framework, which is utilized to detect the environmental anomaly during crop growth. At the first stage, SVM is adopted to classify the data, Spark framework is utilized to solve optimization problem iteratively; at the second stage, GMM is used to do clustering on the classified data respectively. Spark framework is utilized to update the models internationally until stable, during every iteration. Map phase implements the distribution of the sample points to the models. Reduce phase renew the numbers of models and the parameters. Finally, the detection of environmental anomaly is completed by taking advantages of the clustering result. The results show that the proposed approach can be well applied to actual production.