Dual-frequency ground penetrating radar underground target saliency feature fusion method
By designing a semantic feature extraction network and phase constraints, the contradiction between accuracy and depth resolution in ground penetrating radar data fusion in traditional methods is resolved, efficient and accurate feature information fusion is achieved, and the performance of the detection system is improved.
Patent Information
- Application Number
- CN202510918489.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-03
- Publication Date
- 2025-10-10
AI Technical Summary
Traditional methods have difficulty in effectively utilizing the characteristic information in the original echo data when fusing dual-frequency ground penetrating radar data, and the accuracy and quality of image fusion are difficult to guarantee. Especially in complex geological environments, the contradiction between detection depth and resolution is difficult to resolve.
A parallel network architecture combining deep convolutional neural networks with attention mechanism is adopted to design a semantic feature extraction network. Direct wave interference is removed through data preprocessing. The feature information of high-resolution and low-resolution GPR echo data is extracted and fused by combining phase consistency loss and phase change region gradient loss.
It improves the extraction accuracy and fusion quality of feature information, effectively solves the contradiction between detection depth and resolution, and enhances the saliency and accuracy of target features.