Image classification of hyperspectral remote sensing using semi-supervised learning algorithm.

Area of Science:

• Remote Sensing
• Computer Vision
• Machine Learning

Background:

• Hyperspectral images offer rich spectral and spatial data but pose challenges in processing, analysis, and labeling.
• Accurate classification of hyperspectral remote sensing data is crucial for various applications.

Purpose of the Study:

• To develop an effective hyperspectral image classification method leveraging texture features and semi-supervised learning.
• To address the difficulties in sample labeling and improve classification accuracy.

Main Methods:

• Utilized Local Binary Pattern (LBP) to extract spatial texture features from hyperspectral images.
• Implemented a semi-supervised learning approach combining neighborhood information and priority classifier discrimination for pseudo-labeling.
• Employed sparse representation and mixed logistic regression for robust classification.

Main Results:

• The proposed method achieved higher classification accuracy compared to existing approaches.
• Demonstrated stronger timeliness and improved generalization ability on benchmark datasets (Indian Pines, Salinas, Pavia University).

Conclusions:

• The novel classification method effectively utilizes texture features and semi-supervised learning for accurate hyperspectral image analysis.
• The approach offers a promising solution for enhancing the efficiency and accuracy of hyperspectral remote sensing image classification.