Remote Sensing Image Recognition Based on LOG-T-SSA-LSSVM and AE-ELM Network.

Area of Science:

• Computer Science
• Remote Sensing
• Machine Learning

Background:

• Remote sensing image recognition accuracy is affected by diverse working conditions.
• Existing methods often struggle with complex environmental variations.
• Hierarchical strategies offer a potential solution for improved classification.

Purpose of the Study:

• To develop and evaluate a novel hierarchical network for enhanced remote sensing image recognition.
• To improve classification accuracy and recognition performance under different working conditions.
• To compare the proposed method against traditional and existing advanced techniques.

Main Methods:

• A Logistic-T-distribution-Sparrow Search Algorithm-Least Squares Support Vector Machines (LOG-T-SSA-LSSVM) network was proposed for sample classification and parameter optimization.
• An autoencoder-extreme learning machine (AE-ELM) was integrated for data compression and efficient supervised recognition.
• The AE-ELM network was optimized using the sigmoid activation function and 2000 hidden layer neurons.

Main Results:

• The LOG-T-SSA-LSSVM network demonstrated significantly improved classification accuracy on the UCI dataset compared to contrast networks.
• The AE-ELM network showed good recognition performance, particularly with specific configurations.
• The combined AE-ELM based on LOG-T-SSA-LSSVM classification achieved substantial improvements in recognition accuracy over traditional ELM and PSO-ELM networks.

Conclusions:

• The proposed LOG-T-SSA-LSSVM and AE-ELM integration offers a robust solution for remote sensing image recognition.
• This hierarchical approach effectively addresses challenges posed by varying working conditions.
• The findings indicate a significant advancement in recognition accuracy for remote sensing applications.