Balancing Redundancy Reduction and Feature Retention: a Pca-Svm Approach to Hyperspectral Image Classification

Hyperspectral imaging (HSI) has multiple bands that are strongly correlated, which leads to issues such as information redundancy, the curse of dimensionality, and difficulty in data classification. Therefore, to address this, this paper proposes a hyperspectral image classification method that combines Principal Component Analysis (PCA) with Support Vector Machines (SVM). The method first preprocesses the collected hyperspectral images, followed by dimensionality reduction of the preprocessed hyperspectral data using PCA, thereby overcoming issues like information redundancy while retaining the original feature information of the images as much as possible. After reducing the dimensionality of the spectral images, SVM is employed for the final classification and recognition of hyperspectral images. The proposed method is tested on three public datasets: Botswana, Indian Pines and KSC, and the classification accuracy is analyzed under ratios of test sets to training sets. Experimental results show that as the proportion of the training set increases, the OA, AA and Kappa coefficient are significantly improved, and the feature classes become easier to distinguish, which verifies the effectiveness and practicability of the proposed algorithm.

Ҳали таржима қилинмаган