跳到主要导航 跳到搜索 跳到主要内容

An Active Deep Learning Approach for Minimally-Supervised Polsar Image Classification

  • University of Derby
  • Fudan University
  • Xi'an Electronics and Engineering Institute
  • Nanyang Institute of Technology

科研成果: 书/报告/会议事项章节会议稿件同行评审

4 引用 (Scopus)

摘要

Aiming at improving the classification performance with greatly reduced annotation cost, this paper presents an active deep learning approach for minimally-supervised PolSAR image classification, which integrates active learning and fine-tuning convolutional neural network (CNN) into a principled framework. Starting from a CNN trained using a very limited number of labeled pixels, we iteratively and actively select the most informative candidates for annotation, and incrementally fine-tune the CNN by incorporating the newly annotated pixels. Moreover, to boost the performance and robustness of the proposed method, we employ Markov random field to enforce label smoothness, and data augmentation technique to enlarge the training set. Extensive experiments demonstrated that our approach achieved state-of-the-art classification results with significantly reduced annotation cost.

源语言英语
主期刊名2019 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2019 - Proceedings
出版商Institute of Electrical and Electronics Engineers Inc.
3185-3188
页数4
ISBN(电子版)9781538691540
DOI
出版状态已出版 - 7月 2019
活动39th IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2019 - Yokohama, 日本
期限: 28 7月 20192 8月 2019

出版系列

姓名International Geoscience and Remote Sensing Symposium (IGARSS)

会议

会议39th IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2019
国家/地区日本
Yokohama
时期28/07/192/08/19

学术指纹

探究 'An Active Deep Learning Approach for Minimally-Supervised Polsar Image Classification' 的科研主题。它们共同构成独一无二的指纹。

引用此