A new hyperspectral image classification method based on spatial-spectral features.

Area of Science:

• Remote Sensing
• Computer Vision
• Machine Learning

Background:

• Deep learning frameworks show promise in hyperspectral image classification.
• Existing methods often suffer from high complexity and time consumption.
• Traditional methods may overlook crucial local spatial feature correlations.

Purpose of the Study:

• To develop an efficient and accurate hyperspectral image classification method.
• To address the limitations of high model complexity and ignored spatial features in current techniques.
• To extract and fuse spatial-spectral features for improved classification performance.

Main Methods:

• Dimensionality reduction using Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA).
• Extraction of spatial and texture information via 2D Gabor filters on reduced-dimensional data.
• Convolution with random patches to capture spectral features.
• Fusion of spatial and multi-level spectral features for classification using Support Vector Machines (SVM).

Main Results:

• The proposed method achieved high classification accuracies: 98.09% on Indian Pines, 99.64% on Pavia University, and 96.53% on Kennedy Space Center datasets.
• Demonstrated superior performance over other comparison methods on benchmark datasets.
• Validated the effectiveness of the combined Gabor filter and random patch convolution (GRPC) approach.

Conclusions:

• The proposed GRPC method offers a superior approach for hyperspectral image classification.
• The fusion of spatial and spectral features effectively enhances classification accuracy.
• This method provides a computationally efficient and accurate alternative for hyperspectral data analysis.