研究目的
To present a multitemporal technique for multi-class Change Detection (CD) between pairs of images of a satellite image time series, leveraging the temporal consistency of the pixel status to improve CD results.
研究成果
The proposed multitemporal CD technique significantly improves the detection of changes in SAR image time series, as demonstrated by experimental results on flood detection. Future work includes analyzing longer image time series and using automatic thresholding techniques.
研究不足
The approach models changes with a clear temporal signature characterized by step transitions, potentially filtering out noisy changes with inconsistent behavior in the temporal domain.
1：Experimental Design and Method Selection:
The technique involves an iterative algorithm that first derives CD maps between pairs of images using any multi-class CD technique, then performs a post-detection comparison to enforce temporal consistency of changes.
2：Sample Selection and Data Sources:
A multi-temporal C-band dual-polarization SAR dataset from Sentinel-1 was used, including 10 SLC images.
3：List of Experimental Equipment and Materials:
Sentinel-1 SAR images, DEM-based range-doppler terrain correction using SRTM 1ArcSec elevation data.
4：Experimental Procedures and Operational Workflow:
Pre-processing included radiometric calibration, thermal noise removal, TOPSAR deburst, multilooking, speckle-filtering, and terrain correction.
5：Data Analysis Methods:
The log-ratio operator was used for change index computation, with thresholds manually set for change detection.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
