Convolutional Neural Network-Based Cross-Media Semantic Matching and User Adaptive Satisfaction Analysis Model

Area of Science:

• Computer Science
• Remote Sensing
• Artificial Intelligence

Background:

• Convolutional Neural Networks (CNNs) are widely used for hyperspectral image classification.
• Existing CNN methods struggle to capture spatial pose characteristics effectively.
• Principal Component Analysis (PCA) may discard crucial information during dimensionality reduction.

Purpose of the Study:

• To develop a novel hyperspectral image classification model addressing limitations of current CNN approaches.
• To improve the accuracy and efficiency of hyperspectral image classification.
• To enhance the model's ability to capture spatial pose characteristics.

Main Methods:

• Utilized a novel Extension Attribute Profile Feature (EMAP) Stereo Capsule Network Model.
• Incorporated cross-media semantic matching and user adaptive satisfaction analysis.
• Developed a new CNN-based Pan sharpening algorithm for remote sensing images with dilated convolutions.

Main Results:

• The proposed EMAP Stereo Capsule Network Model significantly improves hyperspectral image classification accuracy.
• The model reduces classification time, especially under complex user adaptive satisfaction scenarios.
• The CNN-based Pan sharpening algorithm demonstrates good generalization performance.

Conclusions:

• The EMAP Stereo Capsule Network Model offers a robust solution for hyperspectral image classification.
• The integration of cross-media semantic matching enhances feature representation.
• The study contributes to more accurate and efficient remote sensing image analysis.