关键词: 光呼吸, 光呼吸, 基因, 突变体, 生物学功能

Abstract:

PSO, in combination with SVM algorithms, which is used to optimize the bands of hyperspectral data. Every search result does not necessarily the same, so very many scholars have questions about the reliability of these algorithms. The hyperspectra data, including the samples from Cunninghamia Lanceolata and Pinus massoniana Lamb in their three different life stages (adults, juveniles, infancy), were analyzed by the methods of first derivative, logarithms and normalization respectively. In order to test the reliability of band selected by the PSO-SVM algorithm, the original data and preprocessed data were reduced the dimension by the PSO-SVM algorithm, and then which were classified by Support Vector Machine (SVM), BP neural network, Mahalanobis distance classification method, Fisher classification method and Bayes classification method. The results were as follows: the classification accuracy was the best and more than 90% by Fisher classification method; more than 80% by SVM and BP neural network; the accuracy by Bayes classification method was the worst and all not more than 90%, at a minimum of 43.75%. Meanwhile, during the classification of Cunninghamia Lanceolata and Pinus massoniana Lamb, comparative analysis of algorithms between PSO-SVM and PCA, shown that PSO-SVM was better than PCA. Hence, it was thought that the characteristic bands extracted by PSO-SVM could be better classified by Fisher classification method, SVM or BP neural network methods. Compared with PCA, the PSO-SVM could effectively extract the spectral characteristics, when the data had very small difference.