Scholarlink

Remote Sensing (RS) is a technique used to gather and interpret information from a distance using sensors. It provides critical data by capturing and analyzing sensed information. It plays a crucial role in mapping of LULC (land use and land cover) indispensable, for effective planning and supervision. This study employed a deep neural network to perform SS (Semantic Segmentation) on LISS-III multispectral images, using Fully Convolutional Networks (FCN) based on U-Net. The technique of SS (Semantic Segmentation) entails of assigning every image element in an illustration to a specific class. Three innovative datasets were developed for the research (Dataset 1: 1,470 images, Dataset 2: 13,500 images dataset 3: 960 images), consisting of LISS-III images with four color channels: Blue, Green, Red and near-infrared region. This dataset includes False Colour Composite (FCC) and ground reference mask images. The experiment identifies four distinct classes: Aquatic areas, Plant cover, Untilled land, and Urban settlements. The model is leveraged to train input images of size 256 × 256 × 3, 128 × 128 × 3 and produces an output matrix of size 256 × 256 × 4, 128 × 128 × 4 representing a One-hot vector mask for the classes. The experimental results demonstrated that the FCN classifier is highly effective in detecting land-use and land-cover classes, achieving an overall accuracy (OAA) of 81% for Dataset 1, 84% for Dataset 2, and 77% for Dataset 3.

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

Land Use and Land Cover Analysis on LISS-III Multi-spectral Images Using UNet Deep Learning Model

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Nirav Desai,
  • Akruti Naik

摘要

Remote Sensing (RS) is a technique used to gather and interpret information from a distance using sensors. It provides critical data by capturing and analyzing sensed information. It plays a crucial role in mapping of LULC (land use and land cover) indispensable, for effective planning and supervision. This study employed a deep neural network to perform SS (Semantic Segmentation) on LISS-III multispectral images, using Fully Convolutional Networks (FCN) based on U-Net. The technique of SS (Semantic Segmentation) entails of assigning every image element in an illustration to a specific class. Three innovative datasets were developed for the research (Dataset 1: 1,470 images, Dataset 2: 13,500 images dataset 3: 960 images), consisting of LISS-III images with four color channels: Blue, Green, Red and near-infrared region. This dataset includes False Colour Composite (FCC) and ground reference mask images. The experiment identifies four distinct classes: Aquatic areas, Plant cover, Untilled land, and Urban settlements. The model is leveraged to train input images of size 256 × 256 × 3, 128 × 128 × 3 and produces an output matrix of size 256 × 256 × 4, 128 × 128 × 4 representing a One-hot vector mask for the classes. The experimental results demonstrated that the FCN classifier is highly effective in detecting land-use and land-cover classes, achieving an overall accuracy (OAA) of 81% for Dataset 1, 84% for Dataset 2, and 77% for Dataset 3.