Scholarlink

<p>To improve the accuracy of remote sensing image scene classification under limited sample conditions, a novel classification method by integrating the optimized EfficientNetV2 with the capsule network is proposed. Firstly, a lightweight EfficientNetV2 is used as the feature extraction backbone, incorporating the efficient multi-scale attention mechanism and the multi-scale feature of Res2Net sub-feature groups to enhance the multi-scale feature extraction capability. Secondly, the optimized EfficientNetV2 is combined with the capsule network, leveraging the strengths of the convolutional neural network in feature extraction while utilizing capsule networks to capture object pose information. Finally, the model is fine-tuned using weights pre-trained on large-scale datasets, which effectively enhances performance in scenarios with limited samples. Experimental results demonstrate that the proposed method achieves superior accuracy in remote sensing image scene classification with limited samples.</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

Limited Sample Remote Sensing Image Classification Method Fused with the Optimized EfficientNetV2 and Capsule Network

  • 期刊论文

  • 发表日期：

    2026年6月25日

查看全文
  • Xin Zhang,
  • Linqi Zhang

摘要

To improve the accuracy of remote sensing image scene classification under limited sample conditions, a novel classification method by integrating the optimized EfficientNetV2 with the capsule network is proposed. Firstly, a lightweight EfficientNetV2 is used as the feature extraction backbone, incorporating the efficient multi-scale attention mechanism and the multi-scale feature of Res2Net sub-feature groups to enhance the multi-scale feature extraction capability. Secondly, the optimized EfficientNetV2 is combined with the capsule network, leveraging the strengths of the convolutional neural network in feature extraction while utilizing capsule networks to capture object pose information. Finally, the model is fine-tuned using weights pre-trained on large-scale datasets, which effectively enhances performance in scenarios with limited samples. Experimental results demonstrate that the proposed method achieves superior accuracy in remote sensing image scene classification with limited samples.