Rapid and noninvasive detection methods of seed vigor, an important index to evaluate seed quality, have been the research focus in recent years. In this paper, the detection method of pea seed vigor based on hyperspectral imaging technology was researched. First, the spectral images of different vigor grade samples with artificial aging were captured, and the original spectrum was pretreated with multiple scattering correction. Secondly, SPA and PCA were used to select respective bands. Finally, PLS-DA and LS-SVM model were established to identify the seed vigor of the pea seed, based on the whole band spectrum, the characteristic bands extracted by SPA and PCA respectively. The results showed that PLS-DA and LS-SVM models are effective, but LS-SVM had better performance. Through comparison, the method using full band spectrum was more accurate, the efficiency of method using 5 characteristic bands extracted by PCA was the highest while the way of extracting the representative band by SPA was the most meaningful to this study which achieved similar accuracy to the full band with only 20 bands. The SPA-LS-SVM method afforded the recognition accuracy (100%) for modeling set and validation set used to determine the vigor of pea seeds. The overall results suggest that hyperspectral imaging technology is useful for classification of different vitality pea seeds with non-destructive manner, which can provide a basis for further development of online scoring devices
A Micro-Damage Detection Method of Litchi Fruit Using Hyperspectral Imaging Technology
