Related Experiment Video
Updated: Mar 16, 2026
07:13
Early Detection of Cyanobacterial Blooms and Associated Cyanotoxins using Fast Detection Strategy
Published on: February 25, 2021
4.6K
Genetic Particle Swarm Optimization-Based Feature Selection for Very-High-Resolution Remotely Sensed Imagery Object
Qiang Chen1,2, Yunhao Chen3,4, Weiguo Jiang5,6
1State Key Laboratory of Earth Surface Processes and Resource Ecology, Beijing Normal University, Beijing 100875, China. chenqiang@mail.bnu.edu.cn.
Sensors (Basel, Switzerland)
|August 3, 2016
Summary
This study introduces a Genetic Particle Swarm Optimization (GPSO) algorithm for feature selection in Object-Based Change Detection (OBCD). GPSO improves detection accuracy and convergence speed for high-resolution remote sensing images.
Area of Science:
- Remote Sensing
- Geospatial Analysis
- Computer Vision
Background:
- Object-Based Change Detection (OBCD) in very-high-resolution remote sensing images relies heavily on feature selection for precision and efficiency.
- Abundant features in image objects present an optimization challenge for OBCD.
Purpose of the Study:
- To propose a Genetic Particle Swarm Optimization (GPSO)-based feature selection algorithm for optimizing feature selection in multiple-feature OBCD.
- To evaluate the effectiveness of the proposed GPSO algorithm using the Ratio of Mean to Variance (RMV) as a fitness function.
Main Methods:
- Developed a GPSO algorithm integrated with object-based image analysis for feature selection in OBCD.
- Utilized the Ratio of Mean to Variance (RMV) as the fitness function within the GPSO framework.
- Applied the algorithm to an object-based hybrid multivariate alternative detection model using Worldview-2/3 imagery.
Main Results:
- GPSO demonstrated significantly improved convergence speed and effectively avoided premature convergence compared to other feature selection algorithms.
- The proposed GPSO algorithm achieved superior overall accuracy (84.17%, 83.59%) and Kappa coefficients (0.6771, 0.6314) in OBCD.
- Sensitivity analysis indicated the algorithm's robustness to initial parameters, though feature count and swarm size influenced performance.
Conclusions:
- The GPSO-based feature selection algorithm offers a robust and effective solution for multiple-feature OBCD.
- The RMV function is well-suited as a fitness function for GPSO in this context.
- The method enhances both the speed and accuracy of change detection in very-high-resolution remote sensing data.
