Clustering and application of grain temperature statistical parameters based on the DBSCAN algorithm

Abstract

Reasonable analysis of grain temperature statistical parameters can assist grain depot inspectors to analyze and detect conditions of historical grain reserves. This paper presents a method of using grain temperature statistical parameters to detect grain inventory modes (empty and aeration) based on DBSCAN (Density-based spatial clustering of applications with noise) algorithm. Statistical parameters of grain temperature during normal storage about one year from 27 grain warehouses in China were calculated and clustered with the DBSCAN algorithm. According to the clustering results, parameters analyzing and grain inventory modes detection experiments were conducted. The results of parameters analyzing showed that grain temperature differences between adjacent layers and aggregation ratios of four-layers grain temperatures could be used to detect empty warehouse, and that the recall rate and precision rate reached 100%. The results of parameters analyzing also showed that grain temperature change rate and standard deviation change rate could be used to detect aeration periods, and that the recall rate was about 85.4% and precision rate was about 97.4%.