Detection of Defects in Malus asiatica Nakai Using Hyperspectral Imaging
Hyperspectral imaging technology was employed to detect defects such as rot, bruise and rust in Malus asiatica Nakai. 213 RGB images of samples, including 3 types of damage samples and sound ones, were acquired by hyperspectral imaging system. Spectral data were extracted from the regions of interest (ROI) using ENVI4.7 software. Competitive adaptive reweighted sampling (CARS) and successive projections algorithm (SPA) were used to select characteristic wavelength points. As a result, 11 and 6 characteristic wavelength points were chosen for CARS and SPA respectively. Extreme learning machine (ELM) discrimination model was established based on the spectral data of selected wavebands. The results showed that the accuracy of the SPA-ELM discrimination model was as great as 94.74%. Then, images corresponding to six sensitive bands (532 nm, 563 nm, 611 nm, 676 nm, 812 nm, and 925 nm) selected by SPA were selected for principal components analysis (PCA). Finally, the images of PCA were employed to identify the location and area of a defect’s feature through imaging processing. Through sobel operator and region growing algorithm, the edge and defective feature of 38 Malus asiatica Nakai can be recognized and the detection precision was 92.11%. This study demonstrated that the defects, (rot, bruise, and rust) of Malus asiatica Nakai can be detected in spectral analysis and feature detection in hyperspectral imaging technology, which provides a theoretical reference for the online detection of defects in Malus asiatica Nakai.
KeywordsMalus asiatica Nakai Defects Hyperspectral imaging Detection
This work was supported by the National Natural Science Foundation of China (31271973) and graduate student education innovation project of Shanxi province (2017SY032). The authors are indebted to professor He Yong (College of Biosystems Engineering and Food Science, Zhejiang University, Hangzhou, China) for generously offering the hyperspectral imaging system.
- 1.Xue, J., Zhang, S., Sun, H.: Detection of shelf life of malus asiatica using near-infrared spectroscopy and softening index. Trans. Chin. Soc. Agric. Mach. 44(8), 169–173 (2013)Google Scholar
- 4.Zhang, B., Huang, W., Li, J., et al.: Detection of slight bruises on apples based on hyperspectral imaging and MNF transform. Spectro. Spectral Anal. 34(5), 1367–1372 (2014)Google Scholar
- 5.Zhao, J., Liu, J., Chen, Q., et al.: Detecting subtle bruises on fruits with hyperspectral imaging. Trans. Chin. Soc. Agric. Mach. 39(1), 106–109 (2008)Google Scholar
- 8.Lv, Q., Tang, M., Cai, J.: Detection of hidden bruise on kiwi fruit using hyperspectral imaging and parallelepiped classification. In: 2010 First International Conference on Cellular, Molecular Biology, Biophysics and Bioengineering (CMBB), pp. 309–312 (2010)Google Scholar
- 11.Chen, B., Lu, B., Lu, D.: Parameter optimization of rapeseed oil content model using a miniature near-infrared spectometer. Mod. Food Sci. Technol. 31(8), 286–292 (2015)Google Scholar
- 12.Yang, Y., Zhang, S., Xue, J., et al.: Dynamic discrimination of subtly bruised lang jujubes based on different visible/near-infrared spectral ranges. Mod. Food Sci. Technol. 31(8), 323–328 (2015)Google Scholar
- 15.Abbott, J.A., Lu, R., Upchurch, B.L., et al.: Technologies for non-destructive quality evaluation of fruits and vegetables. Hortic. Rev. 20, 1–120 (1997)Google Scholar