Research On Perimeter Intrusion Detection Technology Based On Ground Sensors

Driven by the rapid development of information reconnaissance technology,sensor signal analysis and detection technology has made a qualitative leap.However,in the process of perimeter intrusion monitoring,it often faces various challenges such as terrain,weather conditions and natural scenery occlusion.Only by the traditional photoelectric sensor can not accurately judge the intrusion,so the vibration sensor is added to the perimeter monitoring system to make up for the function of the traditional photoelectric sensor.In this paper,from the vibration signal preprocessing,feature extraction and classification recognition of three aspects of the perimeter invasion of people and vehicles detection and recognition.Firstly,in view of the trend items generated by the vibration sensor itself and the interference caused by the surrounding environment to the signal during the signal acquisition process,a signal preprocessing scheme of average method combined with wavelet threshold denoising was proposed.In the process of vibration sensor signal acquisition,part of the data will deviate from the baseline and become a trend item,so the average method is used to smooth the signal and remove the trend item.Then,according to the decomposition and reconstruction principle of wavelet threshold de-noising,the over-influence of environmental noise on the signal is reduced.The preprocessing experiment of human and vehicle vibration signals is carried out by combining average method with wavelet threshold denoising method.The experimental results show that the method improves the signal to noise ratio of the signal compared with the traditional algorithm.Then,the pre-processed vibration signals are subjected to feature extraction.The suitability of each feature parameter for the vibration signal is analyzed in the time domain,frequency domain,and time-frequency domain,and the possibility and applicability of the Mel Frequency Cepstrum Coefficient(MFCC)for vibration signal processing is analyzed based on the similarity of the speech signal and vibration signal in terms of non-smoothness.It is finally determined that the signal is first decomposed using the Complementary Ensemble Empirical Mode Decomposition(CEEMD)in time-frequency analysis,and then the marginal spectrum of the signal is obtained by Hilbert transform,and the fast Fourier transform in MFCC is replaced by the above two steps to achieve the effect of optimized feature extraction.The experimental results show that the improved preprocessing algorithm is superior to the traditional algorithm in feature extraction effect.Finally,based on the recognition target of this paper is mainly people and vehicles,the function of the classifier which is frequently used in machine learning is analyzed.Under the consideration of recognition rate and recognition speed,the support vector machine optimized by genetic algorithm is selected to identify and classify the target,and the detection and recognition of the perimeter intrusion target is completed,which makes up for the defect of low target recognition rate of traditional photoelectric sensor.