Research On Signal Analysis Technique Of Bio-photon Emission Of Wheat Kernels With Hidden Insects

Grain storage security has long been a major concern.The detection and prevention of insects in grain is a key problem in the work of grain storage security.How to detect hidden insects in grain is of great research value.In view of the defects of existing detection of hidden insects,this thesis combines biological photon analysis technology with spectrum analysis.We analyze the spectral characteristics of spontaneous biological photon emission signals of infected wheat and normal wheat based on high-order spectral analysis,the spectral distribution characteristics of spontaneous biological photon emission signals of wheat based on CEEMD and Hilbert spectrum are also studied.BP neural network is used to classify and recognize infected wheat and normal wheat,the recognition efficiency is 92.5%.The main work of this thesis is listed as follows:1.First,remove background noise,remove singular value and apply wavelet in the spontaneous biological photon emission signals of wheat.The experimental result shows that the noise in the signal is well suppressed.2.Use the high-order spectral analysis method in the characteristics of spontaneous biological photon emission signal of infected wheat and normal wheat.Results show that the mean,variance and energy of diagonal slice spectrum of infected wheat are higher than that of normal wheat.The energy of horizontal slice spectrum of infected wheat and normal wheat increases first and then decreases with the increase of frequency.The energy of horizontal slice spectrum of infected wheat is higher than that of normal wheat at the same frequency.The phase of the sum of horizontal slice spectrum of normal wheat and infected wheat fluctuates between-2 and 2 at different frequencies.3.The time-frequency distribution of spontaneous biological photon emission signals in infected wheat and normal wheat is illustrated using CEEMD and Hilbert Spectrum.Experiments show that over 95% of the energy of infected wheat and normal wheat distributes in 0-0.1Hz,the SEF and SGF of Hilbert marginal spectrum of infected wheat are higher than that of normal wheat,the proportion of amplitude of Hilbert marginal spectrum of normal wheat is larger than that of infected wheat in the low-frequency stage of 0-0.02 Hz,and the proportion of infected wheat is larger than that of normal wheat in the high-frequency stage of 0.05-0.5Hz.4.Basis on the frequency domain feature and BP neural network,the classification and recognition rate of infected wheat and normal wheat is also studied.Using the frequency domain feature as the input of the classification model,the classification accuracy is 87.15%.Combine the frequency domain feature and time domain feature as the input of the classification model,the classification accuracy reaches 92.5%,the false alarm rate and missing detection rate are 6.67% and 8.33%,respectively.Compared with the previous classification,these methods can better detect whether wheat contains hidden insects.Numerous examples are given to demonstrate the efficiency of combine biological photonics with spectral distribution characteristics to detect of insects in grain.The research results in this thesis have laid a foundation for the establishment of nondestructive detection model for grain insects.