Scholarlink

<p>Efficient classification of remote sensing data is essential for accurate environmental monitoring and decision-making. This study evaluates and enhances two bio-inspired meta-heuristic algorithms —focusing on Ant Colony Optimization (AntClust) and the Firefly Algorithm (FClust)— for satellite image classification, both capable of operating without prior knowledge of class numbers or initial partitions. Experimental results show that AntClust outperforms FClust in adaptability and versatility. To improve FClust, we propose two hybrid methods: K-FClust, integrating k-means to accelerate convergence and boost accuracy, and F-AntClust, combining Firefly and AntClust to refine results. K-FClust improved classification accuracy from 94.83% (baseline FClust) to 96.47%, while F-AntClust achieved 98.80%, representing a 4.18% gain over FClust and a 2.33% gain over K-FClust. These findings demonstrate the effectiveness of hybrid biomimetic approaches in remote sensing classification and highlight their potential for broader intelligent system applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Hybrid meta-heuristic algorithms for remote sensing data clustering

  • 期刊论文

  • 发表日期：

    2026年3月30日

查看全文
  • Seddik Rabhi,
  • Hadria Fizazi

摘要

Efficient classification of remote sensing data is essential for accurate environmental monitoring and decision-making. This study evaluates and enhances two bio-inspired meta-heuristic algorithms —focusing on Ant Colony Optimization (AntClust) and the Firefly Algorithm (FClust)— for satellite image classification, both capable of operating without prior knowledge of class numbers or initial partitions. Experimental results show that AntClust outperforms FClust in adaptability and versatility. To improve FClust, we propose two hybrid methods: K-FClust, integrating k-means to accelerate convergence and boost accuracy, and F-AntClust, combining Firefly and AntClust to refine results. K-FClust improved classification accuracy from 94.83% (baseline FClust) to 96.47%, while F-AntClust achieved 98.80%, representing a 4.18% gain over FClust and a 2.33% gain over K-FClust. These findings demonstrate the effectiveness of hybrid biomimetic approaches in remote sensing classification and highlight their potential for broader intelligent system applications.