研究目的
To reduce data acquisition time and high-level sidelobes in ground-based synthetic aperture radar interferometry by developing a method using dimension-reduced compressive sensing and multiple measurement vectors model for accurate displacement maps.
研究成果
The proposed method effectively reduces data acquisition time and sidelobes, providing more accurate and coherent displacement maps compared to conventional methods. It improves computational efficiency and storage requirements while maintaining high precision, making it suitable for monitoring infrastructures like walls and dams with sparse targets.
研究不足
The method assumes sparse scenes with few strong targets and small displacements (less than half the grid size). It may not estimate displacements of weak targets with high coherence if they are not selected in the support area. The threshold for support area estimation requires trade-offs and depends on scene properties. Undersampling is currently done in postprocessing, not during acquisition.
1：Experimental Design and Method Selection:
The methodology involves a two-stage approach: first, using conventional fast matched filtering-based focusing methods (specifically TDBP) with undersampled data to estimate the supported area of targets, reducing the dimension for sparse reconstruction. Second, applying compressive sensing with a multiple measurement vectors model (using algorithms like OMP or SOMP) to focus raw data from multiple measurements simultaneously, enhancing coherence and precision.
2：Sample Selection and Data Sources:
Real data from two experiments: one monitoring a containing wall with a Ku-band GB-SAR system, and another involving a corner reflector on a dam. Data were acquired with specific numbers of frequencies and azimuth sampling points, with undersampling done in postprocessing.
3：List of Experimental Equipment and Materials:
A Ku-band GB-SAR system (no specific model or brand mentioned), corner reflector, and computational tools for algorithms.
4：Experimental Procedures and Operational Workflow:
For each experiment, raw data were collected, undersampled by random selection of frequencies and positions, processed using TDBP for initial focusing and support area estimation, then applied with CS and MMV models for sparse reconstruction and displacement estimation. Parameters like threshold values and iteration numbers were set based on scene properties.
5：Data Analysis Methods:
Displacement and coherence maps were generated and compared. Performance metrics included number of coherent targets and root-mean-square error of displacement estimates, with statistical analysis over multiple trials.
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
