Scholarlink

<p>Remote sensing image (RSI) scene classification, which targets classifying RSIs into a group of semantic classes based on their content, holds an extensive application across many fields. With the excellent neural network proficiencies of deep neural networks (DNN), RSI scene classification influenced by deep learning (DL) has been attracted significant attention and led to significant advances. The classification of scene is a crucial investigation issue in RSI, which has attracted many analysts. It holds innumerable tasks owing to manifold problems, like the difficulty of RS scenes, class overlap (a scene might cover objects that fit a distant class), and the complexity of obtaining adequately labelled scenes. DL models are achieving a repute in image feature study and achieving innovative performance in RSI scene classification. This study proposes a Mayfly Optimiser Algorithm with Deep Convolutional Neural Network Scene Classification (MFO-DCNNSC) model on RSIs. Initially, Wiener filtering (WF) was used for pre-processing for improving the image quality. To extract features, the Inception-ResNetv2 model is employed. Moreover, the MFO method is utilised for hyperparameter tuning. Finally, the scene classification method was carried out by employing the deep belief network (DBN) technique. The comparison analysis of MFO-DCNNSC model demonstrated better accuracy of 96.19% and 96.03% on the UCM and AID datasets, respectively.</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

Modeling an enhanced Mayfly optimizer algorithm integrated with deep convolutional neural networks for accurate scene classification in remote sensing imagery

  • 期刊论文

  • 发表日期：

    2026年3月17日

查看全文
  • M. Rega,
  • S. Sivakumar

摘要

Remote sensing image (RSI) scene classification, which targets classifying RSIs into a group of semantic classes based on their content, holds an extensive application across many fields. With the excellent neural network proficiencies of deep neural networks (DNN), RSI scene classification influenced by deep learning (DL) has been attracted significant attention and led to significant advances. The classification of scene is a crucial investigation issue in RSI, which has attracted many analysts. It holds innumerable tasks owing to manifold problems, like the difficulty of RS scenes, class overlap (a scene might cover objects that fit a distant class), and the complexity of obtaining adequately labelled scenes. DL models are achieving a repute in image feature study and achieving innovative performance in RSI scene classification. This study proposes a Mayfly Optimiser Algorithm with Deep Convolutional Neural Network Scene Classification (MFO-DCNNSC) model on RSIs. Initially, Wiener filtering (WF) was used for pre-processing for improving the image quality. To extract features, the Inception-ResNetv2 model is employed. Moreover, the MFO method is utilised for hyperparameter tuning. Finally, the scene classification method was carried out by employing the deep belief network (DBN) technique. The comparison analysis of MFO-DCNNSC model demonstrated better accuracy of 96.19% and 96.03% on the UCM and AID datasets, respectively.