ANN-based Multi Classifier for Identification of Perimeter Events

Identification of perimeter events enables smarter perimeter security systems. This paper presents a multi classifier. Support Vector Machine (SVM) and Artificial Neural Network (ANN) are the bottom to build the classifier. The top level employs voting mechanism to identify intrusions, taking time evolution characters into account. In addition, to make the classifier be more self-adaptive, an incremental learning module is introduced. The proposed classifier has been successfully applied to oil and gas pipeline intrusion detection systems. Practical results show that it can distinguish nuisance events from intrusion events at a high rate of 94.86% and for seven kinds of intrusions, the recognition rate is 95.29%, fully satisfies the real application requirement.