|
Change Point Detection (CPD) in InSAR Time Series using MLP, Case study: Europe Continent
|
Arya Fakhri   , Mehran Satar * |
| University of Isfahan |
|
|
Abstract: (203 Views) |
The Earth's surface is constantly changing due to natural phenomena such as earthquakes and volcanic activity, as well as human activities like groundwater extraction and mining. In recent years, increased awareness of the risks associated with ground movements has led to a higher demand for comprehensive and reliable information regarding these displacements. Various methods have been proposed to estimate land surface displacement, among which time-series analysis of displacement derived from InSAR is one of the most important sources for assessing the extent of surface displacements over time. However, the post-processing of time-series data has received less attention. The most efficient post-processing techniques for time-series analysis involve identifying change points in the time series. By detecting these change points, it is possible not only to extract spatial information but also to analyze the temporal aspect of the results. This study explores a method for identifying change points in the time series of land displacements caused by geological and human activities in Europe, utilizing data from the EGMS system. Given the challenges in detecting change points in displacement time series derived from interferometry, such as noise and seasonal behaviors, the use of a multilayer perceptron neural network can be effective in recognizing complex patterns and nonlinear relationships. The results demonstrate an accuracy of over 97% in detecting change points, highlighting the model's ability to identify changes with appropriate precision, which can aid in a better understanding of Earth's dynamics. Furthermore, the capabilities of the proposed model in analyzing displacement data and identifying changes have been evaluated, and the results were compared with conventional methods. Based on this comparison, the proposed method outperforms statistical approaches in terms of accuracy and exhibits approximately 9 times lower computational cost compared to other deep learning methods.
|
|
| Keywords: InSAR, Change point detection (CPD), Time series, MLP, EGMS. |
|
|
|
Type of Study: Research |
Subject:
RS
Received: 2024/12/3 | Accepted: 2025/01/25 | ePublished ahead of print: 2025/02/2
|
|
|
|
|
|
|
| Send email to the article author |
|
|
