基于超像元词包特征和主动学习的高分遥感影像变化检测
杨进一(1993-),男,安徽霍邱人,硕士生,主要从事高分遥感影像信息提取、机器学习研究。E-mail: yangjinyi_gis@163.com |
收稿日期: 2019-03-25
要求修回日期: 2019-07-13
网络出版日期: 2019-10-29
基金资助
国家自然科学基金项目(41801324)
福建省科技厅引导性项目(2017Y0055)
“数字福建”重大项目([2016]203号)
版权
High-Resolution Remote Sensing Imagery Change Detection based on Super-Pixel BOW Features and Active Learning
Received date: 2019-03-25
Request revised date: 2019-07-13
Online published: 2019-10-29
Supported by
National Natural Science Foundation of China(41801324)
Science and Technology Agency of Fujian Province(2017Y0055)
“Digital Fujian” Program([2016]203号)
Copyright
为解决高分辨率遥感影像变化检测中存在底层特征缺乏语义信息、像元级的检测结果存在“椒盐”现象以及监督分类中样本标注自动化程度较低,本文提出一种基于超像元词包特征和主动学习的变化检测方法。首先采用熵率分割算法获取叠加影像的超像元对象;其次提取两期影像像元点对间的邻近相关影像特征(相关度、斜率和截距)和顾及邻域的纹理变化强度特征(均值、方差、同质性和相异性),经线性组合作为像元点对的底层特征;然后基于像元点对底层特征利用BOW模型构建超像元词包特征,并采用一种改进标注策略的主动学习方法从无标记样本池中优选信息量较大的样本,且自动标注样本类别;最后训练分类器模型完成变化检测。通过选用2组不同地区的GF-2影像和Worldview-Ⅱ影像作为数据源进行实验,实验结果中2组数据集的F1分数分别为0.8714、0.8554,正确率分别为0.9148、0.9022,漏检率分别为0.1681、0.1868,误检率分别为0.0852、0.0978。结果表明,该法能有效识别变化区域、提高变化检测精度。此外,传统主动学习方法与改进标注策略的主动学习方法的学习曲线对比显示,改进的标注策略可在较低精度损失下,有效提高样本标注自动化程度。
杨进一 , 徐伟铭 , 王成军 , 翁谦 . 基于超像元词包特征和主动学习的高分遥感影像变化检测[J]. 地球信息科学学报, 2019 , 21(10) : 1594 -1607 . DOI: 10.12082/dqxxkx.2019.190136
The following problems exist in the change detection of high-resolution remote sensing imagery: lack of the semantic information of low-level features, the "salt and pepper" phenomenon in the detection results based on pixel-level methods, and the low degree of sample labeling automation in supervised classification. In this paper, we proposed a change detection method based on super-pixel Bag-of-Words features and active learning. Firstly, we used the entropy rate segmentation algorithm to obtain the segmentation objects of superimposed images. Secondly, we extracted the features of the Neighborhood Correlation Images (correlation, slope, and intercept) and the change intensity features of texture (mean value, variance, homogeneity, and dissimilarity) while considering neighborhood context information between pixel pairs of the studied two phases of images, and then combined them as the low-level features of pixel pairs. Followingly, based on these low-level features, we constructed the expression of Bag-of-Words features in the super-pixel regions by the Bag-of-Words (BOW) model, and we adopted an improved annotation strategy to annotate automatically the samples with large information from the unlabeled sample pool. Finally, we conducted the change detection using the trained classification model. By choosing two groups in different parts of GF-2 imagery and Worldview-Ⅱ imagery as a data source for experiments, the experimental results show that the F1 scores of the two groups of data sets are 0.8714 and 0.8554, the precision is 0.9148 and 0.9022, the missed detection rate is 0.1681 and 0.1868, and the false detection rate is 0.0852 and 0.0978, respectively. The results demonstrate that our proposed method can effectively ide.pngy the variation area, improve the accuracy of change detection. In addition, the comparison of the learning curves between the traditional active learning method and the active learning method with improved annotation strategy shows that the improved annotation strategy can effectively improve the automation degree of sample annotation at a lower precision loss.
表1 混淆矩阵Tab. 1 Confusion matrix |
真实变化像元数 | 真实未变化像元数 | 总和 | |
---|---|---|---|
检测变化像元数 | TP | FP | P |
检测未变化像元数 | FN | TN | N |
总和 | P' | N' | T |
表2 8种基于相同超像元分割结果的变化检测方法的过程描述Tab. 2 Process description of eight change detection methods based on same superpixel segmentation results |
方法 | 过程描述 |
---|---|
BOW-IAL | 基于视觉词典构建样本的词包特征 在ELM训练中采用主动学习采样,并利用改进标注策略自动标注所采样本,完成训练、变化检测 |
BOW-RS | 构建样本的词包特征(同BOW-IAL) 在ELM训练中采用随机方法进行采样,并利用改进的标注策略自动标注所采样本,完成训练、变化检测 |
LOW-IAL | 构建样本的底层特征(前文所述的3种NCIs特征、4种纹理特征,共16维特征),其中超像元NCIs特征构造方法参考文献[28] 在ELM训练中采用主动学习采样,并利用改进标注策略自动标注所采样本,完成训练、变化检测 |
LOW-RS | 构建样本的底层特征(同LOW-IAL) 在ELM训练中采用随机方法进行采样,并利用改进的标注策略自动标注所采样本,完成训练、变化检测 |
BOW-AL | 构建样本的词包特征(同BOW-IAL) 在ELM训练中采用主动学习采样,人工标注所采样本,完成训练、变化检测 |
LOW-AL | 构建样本的底层特征(同LOW-IAL) 在ELM训练中采用主动学习采样,人工标注所采样本,完成训练、变化检测 |
OCVA | 基于超像元分割结果构建光谱变化强度图 利用大津法获取最佳变化阈值,阈值分割得到变化检测结果 |
SVM | 基于超像元分割结果选取光谱和纹理(光谱特征包括光谱均值和方差,纹理特征包括均值、方差、同质性和相异性)构建 差异特征向量 随机选择总样本的60%作为训练集,剩余40%为测试集,训练SVM分类器,完成变化检测 |
表3 数据集1变化检测结果精度评价Tab. 3 Accuracy evaluation of dataset 1 change detection results |
方法 | F1分数 | 正确率 | 漏检率 | 误检率 |
---|---|---|---|---|
BOW-IAL | 0.8714 | 0.9148 | 0.1681 | 0.0852 |
BOW-RS | 0.7623 | 0.8646 | 0.3183 | 0.1354 |
LOW-IAL | 0.8143 | 0.8369 | 0.2072 | 0.1631 |
LOW-RS | 0.6774 | 0.7652 | 0.3924 | 0.2348 |
OCVA | 0.6849 | 0.6332 | 0.2541 | 0.3668 |
SVM | 0.8261 | 0.8455 | 0.1924 | 0.1545 |
表4 数据集2变化检测结果精度评价Tab. 4 Accuracy evaluation of dataset 2 change detection results |
方法 | F1分数 | 正确率 | 漏检率 | 误检率 |
---|---|---|---|---|
BOW-IAL | 0.8554 | 0.9022 | 0.1868 | 0.0978 |
BOW-RS | 0.7383 | 0.8312 | 0.3359 | 0.1688 |
LOW-IAL | 0.7847 | 0.7969 | 0.2272 | 0.2031 |
LOW-RS | 0.6316 | 0.7121 | 0.4326 | 0.2879 |
OCVA | 0.6668 | 0.6230 | 0.2829 | 0.3770 |
SVM | 0.7919 | 0.8009 | 0.2169 | 0.1991 |
[1] |
张良培, 武辰 . 多时相遥感影像变化检测的现状与展望[J]. 测绘学报, 2017,46(10):1447-1459.
[
|
[2] |
|
[3] |
|
[4] |
|
[5] |
|
[6] |
|
[7] |
|
[8] |
|
[9] |
王成军, 毛政元, 徐伟铭 , 等. 超像素与主动学习相结合的遥感影像变化检测方法[J]. 地球信息科学学报, 2018,20(2):235-245.
[
|
[10] |
|
[11] |
翟俊海, 臧立光, 张素芳 . 在线序列主动学习方法[J]. 计算机科学, 2017,44(1):37-41.
[
|
[12] |
|
[13] |
|
[14] |
赵理君, 唐娉, 霍连志 , 等. 图像场景分类中视觉词包模型方法综述[J]. 中国图象图形学报, 2014,19(3):333-343.
[
|
[15] |
|
[16] |
|
[17] |
浮瑶瑶, 柳彬, 张增辉 , 等. 基于词包模型的高分辨率SAR 图像变化检测与分析[J]. 雷达学报, 2014,3(1):101-110.
[
|
[18] |
|
[19] |
罗星, 徐伟铭, 王佳 . 基于对象BOW特征的高分辨率遥感影像变化检测方法[J]. 地球信息科学学报, 2018,20(8):1150-1159.
[
|
[20] |
|
[21] |
|
[22] |
冯文卿, 眭海刚, 涂继辉 , 等. 联合像素级和对象级分析的遥感影像变化检测[J]. 测绘学报, 2017,46(9):1147-1155.
[
|
[23] |
|
[24] |
|
[25] |
|
[26] |
|
[27] |
|
[28] |
|
[29] |
|
[30] |
|
[31] |
徐盛 . 基于主题模型的高空间分辨率遥感影像分类研究[D]. 上海:上海交通大学, 2012.
[
|
[32] |
|
[33] |
陈伟锋, 毛政元, 徐伟铭 , 等. 基于Adaboost的高分遥感影像自动变化检测方法[J]. 地球信息科学学报, 2018,20(12):1756-1767.
[
|
/
〈 |
|
〉 |