Hyperspectral Image Classification with Optimized Compressed Synergic Deep Convolution Neural Network with Aquila...

Area of Science:

• Remote Sensing
• Computer Vision
• Machine Learning

Background:

• Hyperspectral image classification (HSI) is crucial for Earth observation.
• Deep learning methods are increasingly used for HSI classification.
• Accurate classification with limited labeled data remains a challenge.

Purpose of the Study:

• To develop a novel deep-learning-based method for accurate HSI classification using limited samples.
• To integrate spectral and spatial features for enhanced classification.
• To improve training stability, reduce computational cost, and boost classification performance.

Main Methods:

• A novel deep-learning framework for feature extraction and classification.
• Extraction and integration of spectral and spatial information to generate fused features.
• Implementation of a compressed synergic deep convolution neural network with Aquila optimization (CSDCNN-AO) model.

Main Results:

• The CSDCNN-AO model demonstrated superior performance on four benchmark HSI datasets (KSC, IP, HU, SS).
• Outperformed conventional techniques in average accuracy (AA), overall accuracy (OA), and Kappa coefficient (k).
• Significantly reduced training time and computational cost, enhancing training stability and accuracy.

Conclusions:

• The proposed CSDCNN-AO framework effectively classifies hyperspectral images with limited labeled samples.
• The integration of spectral-spatial features and Aquila optimization leads to enhanced classification accuracy and efficiency.
• This approach offers a promising solution for real-world HSI classification applications.