| Hyperspectral technology is adopted in this thesis to detect the rice seed vigor rapidly and nondestructively. Seed vigor is not only one of the important indicators, but the key to the success of security of grain production on behalf of the seed quality. Seed vigor measurement is important in the storage of seed. The traditional detection method of seed vigor is mainly with the aid of artificial to complete which is complicated operation, workload is big and cycle is long, it is also has certain destruction. Therefore, the establishment of seed vigor of rapid, accurate and nondestructive detection method, not only can improve the efficiency and level of seed vigor detection, but also has important meaning for the seed quality management.Previously, seed quality is mainly detected by traditional ways which is not only need more time and more people to complete but has certain damage for seed itself, The more damage is not recycled. Therefore, the establishment of seed vigor of rapid and nondestructive detection method is quite important for the actual production and the efficiency of the seed vigor.Hyperspectral imager for rice seed is used in this thesis to obtain near infrared spectrum. In the process, the principal component analysis-support vector machine (SVM) pattern recognition method., different number of principal components, characteristics of the band and the kernel function of support vector machine (SVM) and the influence of the parameter selection of the model are adopted to establish model of rice seeds of different vigor level. The number of mistakes is the standard according to the model test, the characteristics of the data analysis methods, principal component number and number of wavelengths. Finally, the independent principal component analysis is used to analyse the spectra data selection of characteristic analysis, kernel functions for SVM linear kernel function, The result is that5～7principal components is the best to establish the model of the optimal. |
PDF Full Text Request