Detection Of Internal And External Quality Of Potato Based On Hyperspectral Technology

The realization of online agricultural nondestructive testing is a very effective method for improving the quality of agricultural products and sales efficiency.However,the detection of agricultural products in the actual life is mainly in the manual detection of artificial detection or related machinery.This method is not only inefficient,And may be due to man-made uncontrollable factors lead to the wrong.With the rapid development of hyperspectral technology,more and more research on the use of hyperspectral technology to detect agricultural products,hyperspectral imaging technology with high resolution,the sample without pretreatment operation,simple operation,non-destructive characteristics,these advantages Making hyperspectral technology to detect agricultural products become a hot spot.In this paper,potato was used as the object of study to explore the feasibility of using hyperspectral technique to detect the internal and external quality of potato.Firstly,the external reflection spectrum image and the internal transmission spectrum data of the potato samples were obtained.The influence of different pretreatment methods on the original spectral data was analyzed and the characteristic bands were selected by different methods.Finally,the corresponding quantitative and qualitative models were established to analyze the whole data analysis The rationality.Specific research content and conclusions are as follows:(1)The experimental data of the external reflection spectrum images and the internal transmission data of the pota to samples were set up.The final reflection spectra and transmission spectra were obtained by constantly debugging the relevant parameters.The spectral data of the region of interest were obtained and the range of 450 nm to 951 nm Spectral data are used for subsequent analysis.(2)The principal component analysis of the external reflection spectral images of the four external quality potato samples was carried out.It was found that the difference between the defective parts and the normal parts of t he second principal component image was relatively large,thus determining the second principal component image as the original image of the image processing.The second principal component images of 8 bands,such as 480 nm,550nm,660 nm,680nm,800 nm,850nm,920 nm and 950 nm,are characterized by the weight coefficients of the second principal component,and the second principal component analysis is performed on these feature images,It is found that the difference between the second principal component image and the first principal component analysis is small,so that the second principal component image of the second principal component analysis is subjected to image processing to separate the defective portions,and the result shows that the qualified potato,The defect rate of potato and potato samples was higher,and the segmentation rate of defective potato samples was lower.(3)(MAS),Savitzky-Golay smoothing,median filtering,normalization,first derivative,second derivative,multiple scattering correction(MSC),standard normal variable correction(SNV),(MC)consists of treatments without pretreatment,comparing the differences between the prepared spectral curves and the untreated spectral curves,and selecting the best pretreatment method b y separately establishing the corresponding PLS model.The results show that standard normal variable correction in the corresponding data after the PLS model established the best performance,thus determining the standard normal variable correction for the best pretreatment method.(4)Based on the successive projection algorithm,13 feature bands are selected,and the method of selecting the characteristic band is improved by the traditional principal component analysis.A new weighted weight algorithm is proposed,and nine bands are selected as the characteristic band based on this algorithm.The support vector machine(SVM)discriminant model is established for the spectral data of the characteristic band selected by the above two methods.The result s show that the feature band based on the weighted weight method is superior to the successive projection algorithm.(5)After obtaining the transmission spectrum data of the potato samples,the noise reduction processing was carried out by multiple scattering correction.The algorithm was applied to the successive projection algorithm,the competitive adaptive weighting algorithm and the uninformative variable elimination algorithm respectively by the genetic algorithm and 18,16,53 characteristic band s are selected.The results show that the effect of the genetic algorithm combined with the successive projection algorithm is better than that of the other two methods.The results show that the proposed algorithm is better than the other two methods.