电子学报 ›› 2020, Vol. 48 ›› Issue (1): 75-83.DOI: 10.3969/j.issn.0372-2112.2020.01.009

• 学术论文 • 上一篇    下一篇

基于Word Embedding的遥感影像检测分割

尤洪峰1, 田生伟2, 禹龙3, 吕亚龙1   

  1. 1. 新疆大学信息科学与工程学院, 乌鲁木齐 830046;
    2. 新疆大学软件学院, 乌鲁木齐 830046;
    3. 新疆大学网络中心, 乌鲁木齐 830046
  • 收稿日期:2018-12-04 修回日期:2019-06-05 出版日期:2020-01-25
    • 通讯作者:
    • 田生伟
    • 作者简介:
    • 尤洪峰 男,1991年6月生,江苏省南通人.现为新疆大学硕士研究生,主要研究方向为人工智能和图像处理.E-mail:1053109177@qq.com;禹龙 女,1974年10月出生,新疆乌鲁木齐人.教授、博士生导师.现为网络与信息技术中心网络部主任,主要研究方向为人工智能、网络空间安全、自然语言处理.
    • 基金资助:
    • 新疆维吾尔自治区自然科学基金 (No.2016D01C050); 新疆自治区科技人才培养 (No.QN2016YX0051)

Remote Sensing Image Detection and Segmentation Based on Word Embedding

YOU Hong-feng1, TIAN Sheng-wei2, YU Long3, Lü Ya-long1   

  1. 1. School of Information Science and Engineering, Xinjiang University, Urumqi, Xinjiang 830046, China;
    2. Software College, Xinjiang University, Urumqi, Xinjiang 830046, China;
    3. Network Center, Xinjiang University, Urumqi, Xinjiang 830046, China
  • Received:2018-12-04 Revised:2019-06-05 Online:2020-01-25 Published:2020-01-25
    • Corresponding author:
    • TIAN Sheng-wei
    • Supported by:
    • Natural Science Foundation of Xinjiang Uygur Autonomous Region,  China (No.2016D01C050); Science and Technology Talent Cultivation Project of Xinjiang Uygur Autonomous Region,  China (No.QN2016YX0051)

摘要: 遥感影像检测分割技术通常需提取影像特征并通过深度学习算法挖掘影像的深层特征来实现.然而传统特征(如颜色特征、纹理特征、空间关系特征等)不能充分描述影像语义信息,而单一结构或串联算法无法充分挖掘影像的深层特征和上下文语义信息.针对上述问题,本文通过词嵌入将空间关系特征映射成实数密集向量,与颜色、纹理特征的结合.其次,本文构建基于注意力机制下图卷积网络和独立循环神经网络的遥感影像检测分割并联算法(Attention Graph Convolution Networks and Independently Recurrent Neural Network,ATGIR).该算法首先通过注意力机制对结合后的特征进行概率权重分配;然后利用图卷积网络(GCNs)算法对高权重的特征进一步挖掘并生成方向标签,同时使用独立循环神经网络(IndRNN)算法挖掘影像特征中的上下文信息,最后用Sigmoid分类器完成影像检测分割任务.以胡杨林遥感影像检测分割任务为例,我们验证了提出的特征提取方法和ATGIR算法能有效提升胡杨林检测分割任务的性能.

关键词: 注意力机制, 图卷积网络, 独立循环神经网络, 并联算法, 词嵌入

Abstract: Remote sensing image detection and segmentation technology usually needs to extract image features and mine the deep features of images through deep learning algorithm. However, traditional imaging features (e.g., color, texture, spatial relationship) cannot fully reflect the semantic information of the images,while single/sequential algorithm cannot fully exploit the deep features and the contextual semantic information of the images. Aiming at the above challenges, in this paper, the spatial relation features are mapped into real dense vectors by word embedding, which are combined with color and texture features. Further, we propose a new parallel algorithm referred to as attention graph convolution networks and independently recurrent neural network (ATGIR) based on graph convolution network and independent recurrent neural network under attention mechanism for remote sensing image detection and segmentation. Our algorithm first assigns probabilistic weights to the combined features based on attention mechanism; then extracts deep features based on the features with high weights to generate labels with directions by using graph convolution network (GCNs) algorithms, extracts contextual semantic information of the images by using the independently recurrent neural network (IndRNN) algorithm; finally, our algorithm realizes image detection and segmentation by using Sigmoid.For remote sensing image detection and segmentation of populous euphratica forest as an instance, we prove that our feature extraction method and proposed ATGIR algorithm can effectively improve the detection and segmentation tasks.

Key words: attention mechanism, GCNs, IndRNN, parallel algorithm, word embedding

中图分类号: