Research Of Location Algorithm And Pattern Recognition Technology Based On Dual M-Z Perimeter Protection System

Optical fiber sensing with stable, long-distance distributed measurement, and anti-electromagnetic interference performance, has become a research hotspot in the field of perimeter protection. Depending on the difference in working mechanisms, the fiber perimeter protection system is divided into:interference technique, fiber grating technology and optical time domain reflectometry, et al. Among them, the dual M-Z Interferometer technology perimeter protection system with high sensitivity, demodulation algorithm simple and low cost, has attracted significant attentions. However, There is some difficulties in this technology, including real-time location technology and sensor signal pattern recognition.According to signal detection, wavelet packet analysis and BP neural network theory, this paper investigates the location algorithm and pattern recognition based on dual M-Z perimeter protection system through theory analysis and experimental verification.Firstly, this paper utilizes a FFTW frequency domain cross-correlation algorithm to achieve real-time location. The proposed algorithm utilizes a cross-correlation operation in the frequency domain instead of traditional time domain operation, so that it can reduce the computation time and improve the system real-time location a large extent.Next, in order to improve the stability of the perimeter system and reduce system false alarm rate, the pattern recognition needs to distinguish different types of sensor signals. This paper utilizes a pattern recognition algorithm to solve this problem. In this algorithm, the wavelet packet analysis is applied to extract the sensor signal energy spectrum characteristic in different spectral bands, and a BP neural network is used to realize pattern recognition, improving the recognition accuracy.Finally, this paper carries out a dual M-Z perimeter protection system, which is composed of a dual M-Z interferometer, a photoelectric detection module, data acquisition module and a signal processing module. The effectiveness of the entire system is verified through experimental results. Experimental results show that the FFTW frequency domain cross-correlation algorithm can achieve real-time location, and vibration signal pattern recognition algorithm can effectively achieve sensor signal pattern recognition.