- 文章名称: Automatic Updating of Land Cover Maps in Rapidly Urbanizing Regions by Relational Knowledge Transferring from GlobeLand30
- June 2019Remote Sensing 11(12):1397
DOI: 10.3390/rs11121397
LicenseCC BY
Project: National Natural Science Foundation of China (No.41631176)
-
问题:conventional approaches always have the limitations of large amounts of sample collection and exploitation of relational knowledge between multi-modality remote sensing datasets
-
目标:produce new land-cover maps based on the existing land-cover products and time series images
- 从历史影像中提取置信度高的土地覆盖信息
- 知识迁移
- 训练并分类
-
supporting Global Human Settlement Layer (GHSL)
-
Such shifts due to differences in acquisition and atmospheric conditions or to the changes in landscape are defined as data-shift problems[Tuia, D.; Persello, C.; Bruzzone, L. Domain Adaptation for the Classification of Remote Sensing Data: An Overview of Recent Advances. IEEE Geosci. Remote Sens. Mag. 2016, 4, 41–57. ]
-
TL: TL is defined as follows [35]: given data and the learning task of source domain, data and learning task of target domain, transfer learning aims to help improve the learning of the target predictive function in target domain using the knowledge in source domain and its learning task.
- a widely used approach is based on adaptation of classifier with source domain samples and labeled/unlabeled target-domain samples
- searching a shared and invariant feature subset(find an appropriate subspace by feature selection methods)
- instance transfer, parameter transfer, feature representation transfer and relational knowledge transfer
-
The method was designed to leverage multi-modality remote sensing dataset, capture high-quality sample labels from land-cover product, and derive the updated land-cover map by unsupervised knowledge transfer procedure. In addition, a novel sample selection strategy is designed as an alternative to traditional random selection method.
Function
-
主要方法:
- Minimize differences of land-cover information between X1 and P1 by multi-modality remote sensing dataset, and establish strategy to produce a subset of P1. Accordingly, the subset C1 is considered as a more reliable land-cover map of X1.
- Establish effective method to select better samples rather than randomly selecting from C1 and obtain a reliable training set T1.
- The pixel labels of X1 which has a high probability to be reliable also for X2 is transferred. Then a training set T2 is obtained.
- According to the result of step 3, use spatio-spectral image classification method to output the land-cover classification result of X2.
-
Land-Cover Optimization Based on Decision Rule
-
The development of decision rules integrated the quality-enhanced NTL(The nighttime light) data, the elevation and slope extracted from DSM and a series of indices extracted from Landsat imagery. Indices related to vegetation, water, and impervious surface were calculated for decision rules. The use of the Normalized Difference Vegetation Index (NDVI) is currently accepted to enhance the difference between vegetation and non-vegetation. The Modified Normalized Difference Water Index (MNDWI) can significantly enhance the water information, especially in urban scenes [46]. The Biophysical Composition Index (BCI) has proved to have a closer relationship with impervious surface abundance than other indices [3]. These three indices were used to aid in the formulation of decision rules. 使用指数监测类别异常区
-
阈值确定方式:he Otsu’s binarization algorithm was applied for threshold decision
T = a r g m a x [ ω 0 ω 1 ( μ 1 − μ 0 ) 2 ] T=arg max[ω_0ω_1(μ_1−μ_0)^2] T=argmax[ω0ω1(μ1−μ0)2]
大津法由大津(otsu)于1979年提出,对图像Image,记t为前景与背景的分割阈值,前景点数占图像比例为 w 0 w_0 w0,平均灰度为 u 0 u_0 u0;背景点数占图像比例为 w 1 w_1 w1,平均灰度为 u 1 u_1 u1。图像的总平均灰度为:u=w0u0+w1u1。从最小灰度值到最大灰度值遍历t,当t使得值g=w0*(u0-u)2+w1*(u1-u)2 最大时t即为分割的最佳阈值。对大津法可作如下理解:该式实际上就是类间方差值,阈值t分割出的前景和背景两部分构成了整幅图像,而前景取值u0,概率为 w0,背景取值u1,概率为w1,总均值为u,根据方差的定义即得该式。因方差是灰度分布均匀性的一种度量,方差值越大,说明构成图像的两部分差别越大, 当部分目标错分为背景或部分背景错分为目标都会导致两部分差别变小,因此使类间方差最大的分割意味着错分概率最小。
- Knowledge Transfer Procedure
- (1) perform Spectral Characteristic-based (SC) clustering using the optimized procedure output; 构造二进制特征矩阵进行分类→放大差异
- (2) assign a training set T 1 T_1 T1 using discriminate criterion based on clustering result; 使用JM距离进行判断
- (3) identify changed and unchanged regions, and label unchanged pixels from X2 using transferred land-cover knowledge. 变化检测 欧几里德距离
-
特征提取并分类 RF SVM bagging method
-
特征提取并分类 RF SVM bagging method
论文地址:https://www.researchgate.net/publication/333720715_Automatic_Updating_of_Land_Cover_Maps_in_Rapidly_Urbanizing_Regions_by_Relational_Knowledge_Transferring_from_GlobeLand30