Research And Application Of Neural Network In Fiber Intrusion Detection System Based On DSP

As the security situation is worsening, the intrusion behavior increasingly diversification, the traditional means of security can no longer good to meet people’s demand for security systems. Optical fiber perimeter detection is a new detection technology which develops with optical fiber sensing technology and optical fiber communication technology. Because of its corrosion resistance, high sensitivity, good electrical insulation, good concealment and other advantages which traditional security systems cannot have, the optical fiber perimeter detection has become the inevitable trend of the development of security system in future.Facing the worsening security situation, people put forward higher requirements on the optical detection system. People not only need to know whether the intrusion occurred, but also need to identify the type of intrusion in real-time, thus can take the targeted protective measures in the first time and minimize the loss. In addition, as the intrusion recognition function, is usually completed by general-purpose computer which has a powerful computing capability, it separates the system of photoelectric detection module and data processing module, to a certain extent, it reduces the system Integration and stability, also is not conducive to the development cost control. At the same time, along with the development of instrument intelligent trend, people also raised higher requirement on the equipment miniaturization and portability. Therefore, how to apply the intrusion behavior recognition algorithm in the fiber optic intrusion detection system, thereby separating from the huge industrial machine, make the subject deserves research.To solve the above two problems, this paper take the achievement of intrusion recognition in DSP as the main target, using the neural networks as pattern recognition method, and do a series of studies in the application of neural network in optical fiber intrusion detection system. This paper mainly studied from the following areas. The first chapter introduced the research background and significance. The second chapter introduced the optical fiber sensing technology and the basic principle and composing of the optical fiber intrusion detection system. The third chapter discussed the model identification scheme of fiber optic intrusion detection system, through the analysis of fiber characteristics of vibration signals and system demand for pattern recognition, select method of feature extraction and classification of the judgment, which are the two key steps of pattern recognition. The fourth chapter designed the system pattern recognition algorithm, realized the signal feature extraction based on the wavelet packet decomposition and the intrusion recognition algorithm based on BP neural network, and verified the algorithm through the MATLAB simulation. The fifth chapter realized the verified algorithm in DSP platform. The sixth chapter proposed to improve the algorithm, for the defects of the BP neural network exposed in the experimental process, and validated the improved algorithm effect before and after by experiments. The seventh chapter summarized this thesis, summed up their own work, and put forward the prospects for research.This paper presents neural network in optical fiber intrusion detection system as the classification method to identify multiple types of intrusions, using DSP powerful data processing ability, achieved the whole signal processing algorithms in the DSP. This scheme improves the system integration and stability, and reduces the development cost of the system, while ensuring the excellent intrusion recognition ability of the system at the same time.