地球信息科学学报 ›› 2019, Vol. 21 ›› Issue (11): 1779-1789.doi: 10.12082/dqxxkx.2019.190285

• 遥感科学与应用技术 • 上一篇    下一篇

基于特征压缩激活Unet网络的建筑物提取

刘浩1,2, 骆剑承1,2,*(), 黄波3, 杨海平4, 胡晓东1, 徐楠1,2, 夏列钢4   

  1. 1. 中国科学院遥感与数字地球研究所 遥感科学国家重点实验室,北京 100101
    2. 中国科学院大学,北京 100049
    3. 香港中文大学 地理与资源管理学系,香港 999077
    4. 浙江工业大学 计算机科学与技术学院,杭州 310024
  • 收稿日期:2019-06-09 修回日期:2019-08-12 出版日期:2019-10-25 发布日期:2019-12-11
  • 通讯作者: 骆剑承 E-mail:luojc@radi.ac.cn
  • 作者简介:刘 浩(1995-),男,湖南常德人,硕士生,研究方向为遥感图像信息提取。 E-mail: liuhao@radi.ac.cn
  • 基金资助:
    国家自然科学基金项目(No.41631179);浙江省自然科学基金(No.LQ19D010006);国家重点研发计划项目(No.2017YFB0503600)

Building Extraction based on SE-Unet

LIU Hao1,2, LUO Jiancheng1,2,*(), HUANG Bo3, YANG Haiping4, HU Xiaodong1, XU Nan1,2, XIA Liegang4   

  1. 1. State Key Laboratory of Remote Sensing Science, Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing 100101, China
    2. University of Chinese Academy of Sciences, Beijing 100049, China
    3. Department of Geography and Resource Management, The Chinese University of Hongkong, Hongkong 999077, China
    4. College of Computer Science and Technology, Zhejiang University of Technology, Hangzhou 310024, China
  • Received:2019-06-09 Revised:2019-08-12 Online:2019-10-25 Published:2019-12-11
  • Contact: LUO Jiancheng E-mail:luojc@radi.ac.cn
  • Supported by:
    National Natural Science Foundation of China(No.41631179);Zhejiang Provincial Natural Science Foundation of China(No.LQ19D010006);National Key Research and Development Program of China(No.2017YFB0503600)

摘要:

自动提取城市建筑物对城市规划、防灾避险等行业应用具有重要意义,当前利用高空间分辨率遥感影像进行建筑物提取的卷积神经网络在网络结构和损失函数上都存在提升的空间。本研究提出一种卷积神经网络SE-Unet,以U-Net网络结构为基础,在编码器内使用特征压缩激活模块增加网络特征学习能力,在解码器中复用编码器中相应尺度的特征实现空间信息的恢复;并使用dice和交叉熵函数复合的损失函数进行训练,减轻了建筑物提取任务中的样本不平衡问题。实验采用了Massachusetts建筑物数据集,和SegNet、LinkNet、U-Net等模型进行对比,实验中SE-Unet在准确度、召回率、F1分数和总体精度 4项精度指标中表现最优,分别达到0.8704、0.8496、0.8599、0.9472,在测试影像中对大小各异和形状不规则的建筑物具有更好的识别效果。

关键词: 高空间分辨率遥感影像, Massachusetts建筑物数据集, 建筑物提取, 深度学习, 卷积神经网络, SE-Unet, 损失函数

Abstract:

Automatic extraction of urban buildings has great importance in applications like urban planning and disaster prevention. In this regard, high-resolution remote sensing imagery contain sufficient information and are ideal data for precise extraction. Traditional approaches (excluding visual interpretation) demand researchers to manually design features to describe buildings and distinguishing them from other objects. Unfortunately, the complexity in high-resolution imagery makes these features fragile due to the change of sensors, imaging conditions, and locations. Recently, the convolutional neural networks, which succeeded in many visual applications including image segmentation, were used to extract buildings in high spatial resolution remote sensing imagery and achieved desirable results. However, convolutional neural networks still have much to improve regarding especially network architecture and loss functions. This paper proposed a convolutional neural network SE-Unet. It is based on U-Net architecture and employs squeeze-and-excitation modules in its encoder. The squeeze-and-excitation modules activate useful features and deactivate useless features in an adaptively weighted manner, which can remarkably increase network capacity with only a few extra parameters and memory cost. The decoder of SE-Unet concatenates corresponding features in the encoder to recover spatial information, as the U-Net does. Dice and cross-entropy loss function was applied to train the network and successfully alleviated the sample imbalance problem in building extraction. All experiments were performed on the Massachusetts building dataset for evaluation. Comparing to SegNet, LinkNet, U-Net, and other networks, SE-Unet showed the best results in all evaluation metrics, achieving 0.8704, 0.8496, 0.8599, and 0.9472 in terms of precision, recall, F1-score, and overall accuracy, respectively. Also, SE-Unet presented even better precision in extracting buildings that vary in size and shape. Our findings prove that squeeze-and-excitation modules can effectively strengthen network capability, and that dice and cross-entropy loss function can be useful in other sample imbalanced situations that involve high-resolution remote sensing imagery.

Key words: high spatial resolution remote sensing imagery, Massachusetts building dataset, building extraction, deep learning, convolutional neural network, SE-Unet, loss function