A Fast Two-Dimensional Phase Unwrapping Algorithm Based on Convolutional Neural Network

Two-dimensional phase unwrapping (2-D PU) is the process of converting measured into real in interferometric signal processing. Reliable results are critical for digital elevation model (DEM) generation using synthetic aperture radar (InSAR) and sonar (InSAS). The majority previous research has concentrated on accuracy. Whereas computational efficiency must be taken account measurement system that requires real-time This paper proposes a low-time-consuming algorithm can accomplish high-precision 2-D PU this application scenario. neural network new path-based make up algorithm. First, incorrect region gradient field predicted corrected network. output channelwise variance then calculated used to generate quality maps. Finally, achieve reconstruction, performs path planning flooding integral according maps compensated gradient. also provides recommended data structure implementation ensure algorithm's high efficiency. Experimental InSAR InSAS show proposed highly efficient accurate.