Scholarlink

<p>The task of multiclass change detection focuses on segmenting areas of land change and identifying the type of ground change using bi-temporal remote sensing images. This technology has recently gained prominence for analyzing remote-sensing data, showcasing significant potential for interpreting changes on the Earth’s surface. However, most existing methods are limited to binary change detection, lacking the capability to identify specific change categories. To address this, this research introduces a novel Deep Gated Hyperbolic Sine reinforcement with Q learning (DGHSQ) scheme for multiclass change detection. The process begins with data collection from two sources: the High-Resolution Semantic Change Detection (HRSCD) and the SEmantic Change detectiON Dataset (SECOND). Image quality is enhanced through normalization and histogram equalization techniques. A dual temporal EfficientNet-B7 with a DeepLabV3 + image transformer is then applied to identify significant changes, followed by semantic segmentation using a modified long short-term memory with UNet technique. Finally, DGHSQ is employed for accurately detecting and classifying land cover changes across multiple categories. The results reveal that the proposed technique accomplishes better performance, with overall accuracies of 99.29% and 99.33% on the two datasets, outperforming existing methods.</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

Deep gated hyperbolic sine reinforcement with Q learning based change detection and classification in satellite images

  • 期刊论文

  • 发表日期：

    2025年10月30日

查看全文
  • Jambukeshwar S. Pujari,
  • Javed Wasim,
  • Aprna Tripathi

摘要

The task of multiclass change detection focuses on segmenting areas of land change and identifying the type of ground change using bi-temporal remote sensing images. This technology has recently gained prominence for analyzing remote-sensing data, showcasing significant potential for interpreting changes on the Earth’s surface. However, most existing methods are limited to binary change detection, lacking the capability to identify specific change categories. To address this, this research introduces a novel Deep Gated Hyperbolic Sine reinforcement with Q learning (DGHSQ) scheme for multiclass change detection. The process begins with data collection from two sources: the High-Resolution Semantic Change Detection (HRSCD) and the SEmantic Change detectiON Dataset (SECOND). Image quality is enhanced through normalization and histogram equalization techniques. A dual temporal EfficientNet-B7 with a DeepLabV3 + image transformer is then applied to identify significant changes, followed by semantic segmentation using a modified long short-term memory with UNet technique. Finally, DGHSQ is employed for accurately detecting and classifying land cover changes across multiple categories. The results reveal that the proposed technique accomplishes better performance, with overall accuracies of 99.29% and 99.33% on the two datasets, outperforming existing methods.