Dynamic Image Segmentation based Full-waveform Inversion of Subsurface Penetrating Radar Data

Subsurface penetrating radar (SPR) fullwaveform inversion (FWI) is an emerging near-surface high-resolution imaging technology that quantitatively evaluate the permittivity of the subsurface. However, because echo contains limited information in SPR FWI, the difficulty to obtain accurate inversion result increases. In this work, we develop a FWI algorithm for SPR to improve the performance of SPR FWI. In the optimization, the image segmentation is employed to correct the update direction. Hybrid clustering algorithm including Canopy and K-means were utilized for segmentation. The numerical experiment demonstrates the dynamic image segmentation improve the performance of FWI.