Detection Of Peach Surface Defects Based On Hyperspectral Imaging

Detection of fruit surface defects is an important part of the fruit grading. Peach is used as sample in this study to explore the theory of detecting peach surface defects. The main research contents and conclusions are as follows:(1) Corrected the spectral differences caused by spherical surface. Through Matlab programming implemented respectively4intensity gradience correction methods based on space dimension and spectral dimension by former researcher. And a new correction method based on space dimension was explored in terms of combining the two of them. It was proved the mean normalization is superior.(2) Filtered the band selection methods of hyperspectral image. Employed6band selection algorithms, such as PCA、SI index、JSKF and so on. It is got that band selection method of Mahalanobis distance is superior to elect band combination which can better distinguish brown rot from normal area:Three-band combination of660nm,680nm and700nm can be used to detect brown rot. Scab defect detection can be realized using dual-band combination of745nm and950nm, three-band combination of680nm,730nm and950nm.(3) Explored the segmentation and recognition method detecting brown rot defect on peaches surface. Employed feature angle cosine value Φ(R660,R680,R700)、three-band combination [R660,R680,R700]、dual-band combination [R660,R680]、band ratio R660/R680and band derivative (R660-R680)/d for pixel-based image segmentation on multi-band images, drawn the segmentation result with the eigenvalue of three band combination Φ(R660, R680, R700) was optimal. The recognition rate of brown rot fruit and normal fruit reached96.9%and94.6%, respectively.(4) Explored the segmentation and recognition method detecting scab defect on peaches surface. Making use of the method of dual-band ratio and three-band feature angle cosine value, drawn correct rate of detecting scab are88.4%and87.2%.