Study On Internal Quality Detection Of Kiwifruit Based On Hyperspectral Technology

The quality of the internal quality of kiwifruit directly affects the taste of kiwifruit,and it is also an important indicator for determining the picking time and storage time of kiwifruit.At present,the method for detecting the internal quality of kiwifruit is mainly to use the physical and chemical test method for sampling damage detection.The process is cumbersome and impractical,subjective and can not achieve industrial detection.Therefore,it is important to propose a rapid,accurate and non-destructive method for detecting the internal quality of kiwifruit.This paper mainly tests the soluble solid content,firmness and dry matter content of kiwifruit,which is the difference in the absorption capacity of the spectrally sensitive groups contained in the organic matter of kiwifruit in the hyperspectral.By analyzing and processing,the corresponding quality content of kiwifruit is detected,and then the classification and sales increase the economic benefits,and the planting mode and management process of the fruit farmers can also be guided according to the internal quality content.Detection of the soluble solids content of kiwifruit based on hyperspectral techniques,spectral pretreatment of kiwifruit hyperspectral images by multivariate scatter correction,Using the successive projections algorithm combined with the kernel principal component analysis method to extract the characteristic spectral bands from the preprocessed hyperspectral information,and input it into the trained least squares support vector machine optimized by the particle swarm optimization algorithm to detect the soluble solid content.In this paper,the four characteristic spectral bands extracted by this method are compared with the separate kernel principal component analysis method,competitive adaptive reweighted sampling and kernel principal component analysis method to extract the seven or three characteristic spectral bands respectively.The results show that the root mean square error and correlation coefficient of the training set and test set obtained by this method are 0.2882/0.9103 and 0.3192/0.8936,respectively,and the detection effect is better than the other two.Kiwifruit firmness detection based on hyperspectral technology,spectral pretreatment of hyperspectral information is mainly performed by standard normal variable;In order to extract the hyperspectral information that effectively reflects the firmness of kiwifruit,the synergy interval partial least squares method is used to optimize the optimal combination interval [16 17 19] three sub-intervals,a total of 32 spectral bands,combined with the kernel principal component analysis method for dimensionality reduction.Extracting the first two principal components as characteristic spectral bands input into the trained partial least squares detector for firmness detection,at the same time,the 10 characteristic spectral bands extracted by the interval partial least squares method and the 32 characteristic spectral bands extracted by the synergy interval partial least square method are input into the detector.The results show that the root mean square error and correlation coefficient of the training set and test set obtained by this method are 0.2698/0.9315 and 0.3573/0.8738,respectively,which can effectively detect the firmness of kiwifruit and the detection model is greatly simplified.Kiwifruit dry matter content detection based on hyperspectral technology,spectral pretreatment using multivariate scatter correction,in order to eliminate the redundancy and collinearity of hyperspectral information,five characteristic spectral bands are extracted from the preprocessed hyperspectral information by using the uninformed variable elimination method combined with the successive projections algorithm,and input into the trained least squares support vector machine detection,The results show that the root mean square error and correlation coefficient of the training set and test set obtained by this method are 0.2826/0.9060 and 0.3129/0.8943,respectively,which are suitable for the detection of dry matter content of kiwifruit;The five characteristic spectral bands extracted by the method are compared with the 217 full-spectral bands and the 66 characteristic spectral bands extracted by the uninformed variable elimination method.It is shown that the characteristic spectral band extracted by the uninformed variable elimination method combined with the successive projections algorithm can better reflect the dry matter content of kiwifruit.