Hyperspectral Image Classification Based On Fuzzy Nonparallel Support Vector Machine

Support vector machine has excellent classification performance, and also has a good performance in hyperspectral image classification. However, parallel decision classification planes are difficult to fully satisfy the distribution of real data. In order to solve this problem, some nonparallel support vector machines have been proposed. However, the real data still has the influence of noise, which makes the classification effect poor. This paper proposes a nonparallel support vector machine combined with fuzzy membership, which assigns different weights to different samples according to their importance. It is named Fuzzy Nonparallel Support Vector Machine, FNSVM. By analyzing the experimental results, it is shown that FNSVM achieves better classification results than SVM.