Fully modelling based intrusion discrimination in optical fiber perimeter security system

In order to develop an efficient, accurate and richly functional intrusion discrimination scheme in the optical fiber perimeter security, this paper proposed a fully modelling based scheme for the DMZI vibration system. In this scheme, data modelling is applied in both the feature extraction stage and the pattern classification stage. By means of incorporating the coefficients of AR modelling into the feature vector, the intrusion characteristics are described in a brief, overall and essential way, which helps to recognize more intrusion types than the existing schemes. Moreover, owing to that the sigmoid modelling is applied in the classifier design, the proposed scheme is endowed with a particular function of estimating occurrence probabilities for all intrusion types. Besides, the adoption of AdaBoostSVM technique further enhances the classification rate. Experimental results showed that, with the above techniques incorporated, our proposed discrimination scheme can identify 6 intrusion types with the average classification rate 87.14% using only 4-length feature patterns, which presents vast potentials for DMZI applications.