2018
DOI: 10.3390/s18020559
|View full text |Cite
|
Sign up to set email alerts
|

An Unsupervised Change Detection Method Using Time-Series of PolSAR Images from Radarsat-2 and GaoFen-3

Abstract: The traditional unsupervised change detection methods based on the pixel level can only detect the changes between two different times with same sensor, and the results are easily affected by speckle noise. In this paper, a novel method is proposed to detect change based on time-series data from different sensors. Firstly, the overall difference image of the time-series PolSAR is calculated by omnibus test statistics, and difference images between any two images in different times are acquired by Rj test stati… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

0
8
0

Year Published

2019
2019
2024
2024

Publication Types

Select...
7
1

Relationship

1
7

Authors

Journals

citations
Cited by 16 publications
(8 citation statements)
references
References 32 publications
0
8
0
Order By: Relevance
“…Moreover, the results of dual polarization data based on Weibull (Figure 13b3, b4) or gamma distribution (Figure 13c3,c4) have good performance. However, compared with the results of dual polarization, the results of single-channel SAR data have more omission detection in Figure 13b1(2), c1 (2). This proves that using more polarization channel information can obtain more accurate change detection results.…”
Section: Discussionmentioning
confidence: 81%
See 1 more Smart Citation
“…Moreover, the results of dual polarization data based on Weibull (Figure 13b3, b4) or gamma distribution (Figure 13c3,c4) have good performance. However, compared with the results of dual polarization, the results of single-channel SAR data have more omission detection in Figure 13b1(2), c1 (2). This proves that using more polarization channel information can obtain more accurate change detection results.…”
Section: Discussionmentioning
confidence: 81%
“…On the other hand, as a result of the development of satellite systems, a huge number of remote sensing images can now be acquired to detect these changes. Owing to the explosive increase in remote sensing data, how to detect changes accurately is an active research topic [2]. In this context, multi-temporal remote sensing images of the same region proved particularly useful in the applications of change detection, including urban planning [3,4], agricultural research [5,6], disaster detection [7,8], and wetland detection [9,10].…”
Section: Introductionmentioning
confidence: 99%
“…Furthermore, in addition to the feature itself, other dimensions also have the possibility for clustering. For example, the multi-modal Gaussian modeling method [69] clusters the distances between parameter vectors, the multivariate Gaussian mixed model [113,114] generates unsupervised thresholds for negative change, positive change, and no-change situation.…”
Section: Methods Of Feature Clusteringmentioning
confidence: 99%
“…In addition, with the purpose to optimize change results, performing statistics on the raw difference results of multi-temporal images also plays an important role. Experiences reveal that it is indispensable to iterate model and optimize difference image (DI) by generalized statistical region merging (GSRM), Gaussian Mixture Model (GMM), or other optimized technology [69]. Thereinto, two points are mainly emphasized: one is to improve the completeness of change extraction by repeatedly modeling the difference image (DI), or by repeatedly testing the change with correlation statistics [70].…”
mentioning
confidence: 99%
“…Statistical-based methods use region-level statistical information to build change features, which are then optimized iteratively by statistical probability theory to get the final result. Traditional models include Gaussian mixture model (GMM) and generalized statistical region merging (GSRM) [4] are generally adopted. Feature classification-based methods utilize feature mapping (e.g., support vector machine), dimension reduction (e.g., principal component analysis), and ensemble learning (e.g., decision tree) to predict the feature classification results.…”
Section: Change Detectionmentioning
confidence: 99%