Optimized kernel Nonparametric Weighted Feature Extraction for Hyperspectral Image Classification

Hyperspectral image (HSI) classification is an essential means of the analysis remotely sensed images. Remote sensing natural resources, astronomy, medicine, agriculture, food health, and many other applications are examples possible this technique. Since hyperspectral images contain redundant measurements, it crucial to identify a subset efficient features for modeling classes. Kernel-based methods widely used in field. In paper, we introduce new kernel-based method that defines Hyperplane more optimally than previous methods. The presence noise data HSI causes changes boundary samples and, as result, incorrect class hyperplane training. We propose optimized kernel non-parametric weighted feature extraction classification. KNWFE method, which has promising results classifying remotely-sensed data. However, does not take closeness or distance target Solving problem, KNWFE, better performance. Our extensive experiments show proposed improves accuracy superior state-of-the-art HIS classifiers.