Surface Damage Detection Of Korla Fragrant Pear Based On Spectroscopy And Hyperspectral Image

Pear is easy to be collided and squeezed during picking,storage,and transportation,resulting in large area decay.In this paper,hyperspectral image technology was used to accurately and efficiently detect the surface damage defects of pear.80 fragrant pears were ultilized in this work,and the hyperspectral images of intact samples and damaged samples in the range of 400~1000 nm were collected.Then,the surface damage identification methods based on spectral information and the hyperspectral image were studied respectively.(1)The principle of detecting the surface damage of Korla pear employing visible/near-infrared spectroscopy and hyperspectral image was studied.A variety of algorithms were analyzed,including the wavelength selection methods,the identification models of the surface damage of pear,the spectral ratio of the hyperspectral image for image enhancement,and the segmentation method of the surface damage images.The hyperspectral images of 80 fragrant pear samples(400-1000 nm)were collected by hyperspectral image acquisition equipment,and a small sample set was established.(2)Regions of interests(ROI)were extracted from the collected hyperspectral images of fragrant pears.The average reflectance spectra of intact fragrant pear samples and damaged fragrant pear samples were obtained.SPA,CARS,and RELIEF were used to reduce the dimension of the average reflectance spectrum of fragrant pear.As a result,17,17 and 18 characteristic wavelengths were selected respectively.Using these characteristic wavelengths,the damage detection models of fragrant pear surface based on LDA algorithm,SVM algorithm,and ELM algorithm were constructed respectively.By comparing and analyzing the experimental results,the RELIEF-ELM model can achieve the best detection results.The accuracy was 92.5%.The precision was 94.7% and the recall rate was 90.0%.This method has a high application potential.(3)The method of surface damage detection of Fragrant Pear based on hyperspectral image was studied.Using a statistical analysis method,the hyperspectral image at 863 nm was selected to build the mask image.PCA analysis was used to reduce the dimension of hyperspectral data of fragrant pear,and the PC2 image with the most obvious difference between damage and background area was selected.Band ratio processing was carried out with PC4 to further enhance the difference between the damaged area and the background area.Finally,the difference between the damaged area and the backgroundarea was adapted.Threshold segmentation,morphological open and close operations were used to extract damaged areas on the surface of fragrant pear.The results show that the method can effectively identify the surface damage of fragrant pear.The accuracy,precision and recall rate were 93.75%,87.50%,and 100%,respectively.