Detection Of Fresh Jujube (Li Jujube) Crack Based On Hyperspectral Imaging Technology

In recent years, fresh jujube has got a large production with the extending of market demand. Grading by quality is the vital technology for preservation in storage of the fresh jujube industrialization, which is also the important guarantee to price by the quality of jujube. The paper research the potential of rapid and non-destructive detection for surface crack of fresh jujube based on the hyperspectral image information in visible and near infrared region (380-1030nm).The main contents and conclusions are as follows:(1) In order to reduce the high dimensionality of the hyperspectral information, extract6,5and5sensitive wavebands(SW) of crack of fresh pear jujube by partial least squares (PLS), successive projection algorithm (SPA), uninformative variables elimination (UVE) and method combined with UVE and SPA (UVEPLS-SPA) respectively from the hyperspectral image data gotten by ENVI software.(2) Established BP Neural Network (ANN) and least squares support vector machine (LS-SVM) discriminant models based on the sensitive wavebands and predicted samples of prediction set. It proved that the BP Neural Network model based on sensitive wavebands extracted by PLS was best.(3) Through the analysis for hyperspectral image data, realized the detection of the crack of fresh jujube. Three image processing methods were used as follows:the first, adopted the principal component analysis (PCA) to compress the spectral data and selected the PC6to display the crack best; the second, three SWs (418nm,667nm and731nm) were selected using weighted regression coefficient of PC6and processed the second PCA, the third, processed the PCA to the spectral data twice. It proved that the pc6of the second PCA was best for the detection of crack of fresh jujube.(4) Detection of the crack region was realized by the calculation of crack image. Covering film, binary calculations and image multiplying were adopted to calculate a crack region and the occupied area ratio of spectral based on the pc6of the second PCA.