Research And Implementation Of Optical Fiber Perimeter System Pattern Recognition System

With the rapid development of economy, modern science and technology,the demand for regional security continues to grow. Fiber optic sensor is no longer a bottleneck in engineering applications. Optical fiber perimeter system based on optical fiber vibration sensor become a hotpot. Now the rate of optical fiber perimeter system is high and the detecting efficiency. However,it could be work after a large number time of debugging. This means that virtually increased the system cost. After optical fiber perimeter system is completed in the security area, there is a serious problem how the system can learn and process events, which never encountered. When types of intrusion occur in the sensing area at the same time, the system can not recognize the types of intrusion events. To solve the above problems, this paper proposes a new signal extraction and recognition algorithms which based on optical fiber perimeter system of intervention vibration sensor. including the following aspects:In a complex environment, the system signal extraction method has not established a theoretical basis. In my paper, we first analyze the characteristics of human intrusion events. And this research analyzes and formulates data extraction system based on human engineering. We reformulate the sampling period rules to extract signal. Result of simulation and experiment show that the accuracy rate of intrusion events extraction is 100%. Besides, the data extraction system reduces the probability of system making multiple alarms at one intrusion event. Time-domain signal via FFT obtains frequency-domain signal. Then this paper analyzes the different events signal in time domain and frequency domain waveform characteristics.Event frequency domain feature is used as the main feature domain of event recognition system. These features are extracted by method of principal component analysis (PCA). And the principle and realization method of PCA is expounded. The performance of support vector machine(SVM) and K-means clusteringalgorithm in the system is verified. We analyze unsupervised learning and supervised learning algorithm and discusse the influence of recognition on the optical fiber perimeter system.Identify the differences in the environment, this paper Combine with reinforcement learning theory and known event sample characteristics to improve the generalization of the system.