A structure-tensor-constrained cooperative inversion using an extracted physical property distribution. Issue 2 (11th February 2019)

• Record Type: Journal Article
• Authors: Kim, B; Seol, S J; Byun, J; Cho, S
• Is Part Of: Geophysical journal international. Volume 217:Issue 2(2019)
• Page Start: 1334
• Page End: 1352
• Publication Date: 2019-02-11
• Subjects: Controlled source electromagnetics (CSEM) -- Image processing -- Joint inversion -- Acoustic properties; Geophysics -- Periodicals; 550
• DOI: 10.1093/gji/ggz082
• Languages: English
• ISSNs: 0956-540X

Abstract

SUMMARY: When performing a cooperative inversion using a structural constraint, extracting an accurate average direction from the high-resolution model is important because two models with different resolutions should be used in the procedure. However, when the average direction of the high-resolution model, which indicates the major change direction of the model at each inversion block location, is calculated, the conventional gradient methods such as cross-gradient have a limitation on components having opposite directions. Therefore, in this study, an effective average-direction extraction algorithm was developed by introducing the concept of the structure tensor to accurately calculate average-direction information. And finally, based on the extracted average-direction information, structure-tensor-constrained cooperative inversion algorithm was proposed. To verify the effectiveness of the proposed inversion method, the method was tested using data obtained from the synthetic model containing an anomalous body with a complicated shape and the result compared with results of individual EM inversion and the conventional cross-gradient-constrained cooperative inversion. Lastly, to evaluate the performance with a realistic mode, the proposed cooperative inversion was applied to the data acquired using the complex SEG Advanced Modelling Program model. In all experiments, the cooperative inversion with the structure-tensor constraint provided better location estimation results as well as better estimations of the shape of the anomaly. In addition, the resistivity distribution of the anomaly was estimated to be closer to the truth in the inversion result.