FL-Net Fringe Line Detection and Phase Unwrapping in InSAR

The patent replaces traditional mechanical/mathematical fringe line detection algorithms with a deep learning-based convolutional neural network system. The neural network automatically learns fringe line patterns from training data and performs detection without relying on hand-crafted algorithms, thereby improving robustness against noise while maintaining detection capability.

The patent implements preliminary training of the convolutional neural network using labeled fringe line data before actual detection. This preliminary action allows the system to learn optimal detection features and patterns in advance, enabling it to handle noisy conditions effectively during the actual phase unwrapping process.

The computationally intensive training phase is performed in advance during system setup, allowing the model to learn optimal detection parameters. During actual runtime, the trained model performs rapid inference on new data, separating the heavy computational burden from the time-critical detection process.

The patent uses a trained neural network model that can be copied and deployed multiple times. Once trained on comprehensive data, the model serves as a reusable component that provides consistent, noise-robust detection without requiring retraining for each new dataset, significantly reducing processing time for subsequent operations.