| Grain which is an important strategic material has close relationship with national security and people’s life. Thus the safe and efficient management of grain storage traditionally have been paid much attention in all countries. China’s grain output is 4500 ~500 billion kilograms every year. The loss caused by pest is 45 ~ 15 billion kilograms. It forms the main part of the grain storage loss. Therefore, pest problems should be taken attention at any time during the whole process of food storage. If not, a slight negligence can lead to devastating losses on account of the high-speed reproductive of pests. With the rapid development of semiconductor materials’ science and ultrafast laser technology,the terahertz(THz) spectrum system can provide new biometric information of stored grain pests and the bases of identifying stored grain pests in the terahertz band. It helps to establish a new method to identify and study stored grain insects by mixing other detection techniques′ advantages together. So THz systems can inspect stored grain pests in a convenient and quick way. Therefore it has the important practical significance to investigate, monitor pests and accelerate the accurate detection of pests in the grain depot.Grains and wheats are chosen as a represented example in this thesis. Then we pay attention to the THz time domain spectrums, absorption spectrums, reflection spectrums and frequency spectrums of the weevils, grains and wheats. They make contribution to provide preliminary experimental data for the fingerprint libraries of pests in THz wave band.We try to seek a kind of detection method of pests based on THz waves mentioned above. As thus, we establish analysis detecting models of pests based on terahertz information.In this paper, the main contents and innovative points are as follows:( 1) The paper introduces the research’s background and development, mainly includes two aspects. One is the present situation of detection methods used for the grain insects at home and abroad, analysis of the characteristics of terahertz science and technology, the status and progress in agricultural products and food detection and the importance and significance of researching the terahertz time-domain spectroscopydetection of pests; The other is introducing the typical system of THz, the terahertz transmission system in this experiment, optical parameters extraction principle, the evaluation of signal-to-noise ratio and dynamic characteristics of the system in detail.(2)When analysis spectrum of these three kinds of pure samples which contain corn powder, chaff powder, rice flour respectively, we found they have substantial signal attenuation and different time delay in the time domain graphs. Analysis the absorption coefficient spectrum of the samples, we found that their absorption spectrums have the peak in 0.28 THz, only corn has the apparent absorption peak in 1.46 THz of absorption spectrum which is different from other two kinds. Therefore, through the time delay of time domain graph and the characteristic absorption peaks of absorption spectrum we can judge whether the samples contain corn weevils.(3)Through analyzing these spectrums of 8 kinds of samples, which is mixed with rice and corn. The samples′ proportion of insect powder content is 0%, 15%, 25%, 35%,45%, 55%, 45% and 75% separately. We found the rule that in frequency domain the peak value increases with higher proportion of insect powder content. It shows that the rice noodles′ absorption of terahertz spectrum is better than corn weevil powders′. The above analysis results are consistent with the 3 kinds of pure samples. When analyzing the absorption coefficient spectrum of 8 kinds of samples, we found that they all have a absorption peak in 0.28 THz and an obvious absorption peak in 1.36 ~ 1.36 THz which is due to the existence of corn weevils. Therefore it can be used to verdict if there is weevil in samples.(4)Based on the analysis of the time domain spectral of the 3 kinds of pure samples made up of corn powder, chaff powder, rice flour by HHT, we found that the energy in1.5 ~ 0.5 ps corresponding to 0.5 ~ 1.8 THz and 1.5 ~ 0.5 ps corresponding to 0.5 ~ 0.75 THz areas can be used as rice noodles′ characteristics, while the energy spectra in- 5 ~1.5 ps areas can be used as characteristic spectra of chaff powder. So it is fast and convenient to use a single time-frequency diagram to identify three kinds of pure samples instead of putting the three samples′ time domain or frequency domain figure together.Through analyzing the 8 kinds of mixed samples, we discover that when the content of rice flour is larger, the rice noodles′ feature information in the HHT spectra is significant,meanwhile when corn weevils powder content is larger, the characteristics of corn weevils powder in the HHT spectra is significant as well as. The rules are consistent with pure sample analysis.( 5) By using partial least squares discriminant analysis(PLS- DA) method to establish and identify model to identify whether the sample contains corn weevils composition, the accuracy is as high as 100%. Using PCA(principal component analysis)combining with the method of the k-NN, the total recognition rate of 0.9322. Using PLSR(partial least-squares regression method) to establish model to predict and validate the different fragments contention in the mixed samples with corn weevil composition, the correction model and verification model prediction of internal cross-validation mean square deviation RMSECV are 3.7389% and 3.9664% respectively, meanwhile the correlation coefficient are 0.9985 and 0.9985 separately. It means that the model prediction accuracy is high, stable and reliable. |
PDF Full Text Request