• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>Asian conference on remote sensing >Accuracy Analysis of using Intersection over Union at Normalized Difference Vegetation Index
【24h】

Accuracy Analysis of using Intersection over Union at Normalized Difference Vegetation Index

机译：归一化植被指数下联合点交叉口使用精度分析

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

In South Korea, it is very important to build accurate databases of forests, which occupy most of the territory, in order to prevent disasters and accidents. This paper is about that comparative analysis of generalized intersection over union and error matrix in NDVI. The Normalized Difference Vegetation Index (NDVI) depends largely on the image classification and accuracy setting method. In recent remote exploration, image classification is being carried out based on deep learning. Currently, almost all remote sensing uses error matrix for accuracy analysis, but in the field of image processing in computer science, IoU is used for accuracy analysis. Therefore, the accuracy is analyzed using the IoU method, but the difference in accuracy levels between the Vegetation Index identified using the NDVI and that identified using the deep learning technique cannot be known. Intersection over Union (IoU) is the most popular evaluation metric used in the object detection benchmarks. IoU, also known as Jaccard index, is the most commonly used metric for comparing the similarity between two arbitrary shapes. Therefore, to solve this problem, in this study, the Vegetation Index was calculated using the deep learning method and the accuracy of the Vegetation Index calculated using the existing error matrix and that using the IoU method were compared. To extract objects from satellite images, conifers and deciduous trees were identified using the semantic image classification technique in the CNN(Convolutional Neural Networks) technique using a DeepLab model. DeepLab models are based on atrous convolution. These models have the advantage of easy image classification since their receptive field is wide and carry out calculations as if unpooling and convolution were combined. To evaluate the performance of the model, the accuracy of classification was calculated using IoU, and the results were shown.
机译：在韩国，为了防止灾难和事故，建立精确的森林数据库非常重要，因为该数据库占据了大部分领土。本文是关于NDVI中广义联合与误差矩阵相交的比较分析。归一化植被指数（NDVI）在很大程度上取决于图像分类和精度设置方法。在最近的远程探索中，正在基于深度学习来进行图像分类。当前，几乎所有的遥感技术都使用误差矩阵进行精度分析，但是在计算机科学的图像处理领域，IoU用于精度分析。因此，使用IoU方法分析了精度，但是使用NDVI识别的植被指数和使用深度学习技术识别的植被指数之间的精度水平差异是未知的。联合交叉口（IoU）是对象检测基准中使用的最受欢迎的评估指标。 IoU，也称为Jaccard索引，是比较两个任意形状之间相似度的最常用指标。因此，为解决这个问题，在本研究中，使用深度学习方法计算了植被指数，并比较了使用现有误差矩阵和使用IoU方法计算的植被指数的准确性。为了从卫星图像中提取对象，使用DeepLab模型在CNN（卷积神经网络）技术中使用语义图像分类技术来识别针叶树和落叶树。 DeepLab模型基于无规则卷积。这些模型的优点是易于图像分类，因为它们的接收范围很广，并且可以像合并非卷积和卷积一样执行计算。为了评估模型的性能，使用IoU计算分类的准确性，并显示结果。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号