Abstract:In the domain of real-world concept learning, feature selection plays a crucial role in accelerating learning processes and enhancing the quality of classification concepts, particularly in remote sensing image classification. Traditional methods often exhibit weaknesses such as instability, sensitivity to noise, and a lack of consideration for feature dependencies when selecting informative spectral features from hyper-spectral (HS) images. Recent studies have recommended the utilization of meta-heuristic algorithms to address these limitations. However, most of these approaches overlook the correlation between spectral bands. To tackle these challenges, we propose a novel migration-based particle swarm optimization (MBPSO) algorithm that surpasses previous methods in terms of exploitation and exploration and considers spectral band correlation to prevent the inclusion of unnecessary bands. The variance-based J1 criteria are employed as the fitness function for MBPSO, while a support vector machine (SVM) is utilized as the classifier. To evaluate the performance of our proposed method against other optimization algorithms, we conducted experiments using four HS datasets, namely Indian Pine, Pavia University, Salinas, and Washington Mall. Numerical evaluations encompassing class accuracy, overall accuracy, kappa coefficient, execution time, stability, Wilcoxon test, and convergence rate were performed. The results indicate that our proposed algorithm achieves superior numerical outcomes compared to other algorithms, while requiring fewer features to attain higher accuracy. The algorithm's higher speed, reduced computational time, and avoidance of additional features in the solution can be attributed to the presence of the conditional mutation operator within the MBPSO algorithm.

Classification of Hyperspectral Image Based on SVM Optimized by a New Particle Swarm Optimization

SVM KERNEL PARAMETER SELECTION BASED ON IMPROVED GSO

A Method of Particle Swarm Optimized SVM Hyper-spectral Remote Sensing Image Classification

An Improved Multi-class Support Vector Machine for Hyperspectral Remote Sensing Imagery Classification

Hyperspectral Image Classification Based on Multiple Improved Particle Swarm Cooperative Optimization and Svm

A Band Selection Method for Hyperspectral Image Classification Based on Improved Particle Swarm Optimization

Remote Sensing Image Classification Based on Dynamic Co-evolutionary Parameter Optimization of SVM

Classification of Hyperspectral Image with Feature Selection and Parameter Estimation

Application of improved PSO-SVM approach in image classification

New Method Based on Support Vector Machine in Classification for Hyperspectral Data

Application of SVM Algorithm Based on Particle Swarm Optimization in Data Classification

Subpixel Mapping Method of Hyperspectral Images Based on Modified Binary Quantum Particle Swarm Optimization

A support vector machine based on an improved particle swarm optimization algorithm and its application

Classification of Hyperspectral Remote Sensing Image Using Improved LS-SVM

Semisupervised Hyperspectral Image Classification with SVM and PSO

Wetland Remote Sensing Classification Using Support Vector Machine Optimized with Co-Evolutionary Algorithm

Remote sensing image classification with parameter optimized support vector machine based on evolutionary computation

Towards an Optimal Support Vector Machine Classifier Using a Parallel Particle Swarm Optimization Strategy.

Optimal Band Selection Using Evolutionary Machine Learning to Improve the Accuracy of Hyper-spectral Images Classification: a Novel Migration-Based Particle Swarm Optimization

A Hybrid Particle Swarm Optimization Algorithm with Dynamic Adjustment of Inertia Weight Based on a New Feature Selection Method to Optimize SVM Parameters