Download PDFOpen PDF in browser
EN
The title and the abstract of this preprint are also available in English

结合目标局部和全局特征的CV遥感图像分割模型

EasyChair Preprint no. 3646

12 pagesDate: June 19, 2020

Abstract

随着遥感卫星技术的发展,高分辨率遥感影像不断涌现。从含有较多信息、背景复杂的遥感影像中自动提取目标成为一个亟待解决的难题。传统的图像分割方法主要依赖图像光谱、纹理等底层特征,容易受到图像中遮挡和阴影等的干扰。为此,针对特定的目标类型,提出结合目标局部和全局特征的CV(Chan Vest)遥感图像目标分割模型,首先,采用深度学习生成模型--卷积受限玻尔兹曼机建模表征目标全局形状特征,以及重建目标形状;其次,利用Canny算子提取目标边缘信息,经过符号距离变换得到综合了局部边缘和全局形状信息的约束项;最终,以CV模型为图像目标分割模型,增加新的约束项得到结合目标局部和全局特征的CV(Chan Vest)遥感图像分割模型。在遥感小数据集Levir-oil drum、Levir-ship和Levir-airplane上的实验结果表明:所提模型不仅可以克服CV模型对噪声敏感的缺点,且在训练数据有限、目标尺寸较小、遮挡及背景复杂的情况下依然能完整、精确地分割出目标。

Keyphrases: Convolutional restricted Boltzmann machine, CV model, deep learning, image segmentation, Shape prior

BibTeX entry
BibTeX does not have the right entry for preprints. This is a hack for producing the correct reference:
@Booklet{EasyChair:3646,
  author = {Xiaohui Li and Xili Wang},
  title = {结合目标局部和全局特征的CV遥感图像分割模型},
  howpublished = {EasyChair Preprint no. 3646},

  year = {EasyChair, 2020}}
Download PDFOpen PDF in browser