Multi-target recognition used in airpoty fiber fence warning system

For the shortage of an ultra-long-distance auto-alarming system at the airports, a new fiber optic perimeter security system based on the M-Z interferometers is proposed. This system can be used to identify the invasion of walkers, little animals and vehicles within BP neural network. Firstly, the input signal was preprocessed to select the vibration signals which includes intrusion and takes wavelet transformed in order to remove the noise with appropriate threshold. Secondly, the feature is extracted within wavelet packet. Finally, the invasion behavior which often occurs is identified within BP neural network. Experiments show that, in the complex working environments, this system is able to identify intrusions with an accuracy rate more than 96.8%.