Abstract:In order to solve the problem that the detection of internal fringes deviates from the fringes boundary during the phase unwrapping of GVF Snake fringes detection, a phase unwrapping method combining GVF Snake fringes detection and Markov random field (MRF) graph cutting is proposed. By judging that the fringe detection line is not the real fringe boundary, the inner block of the one-step fringe detection line is segmented and unwrapped by MRF method, and the two unwrapped results are fused according to the corresponding unwrapped criteria to form the rough unwrapped results with boundary jump points. The method of eliminating the boundary isolated points of the rough unwrapped results by high pass filter interpolation is studied. Phase unwrapping in the area with large gradient in the mining area. The results show that the validity of the algorithm is verified by the real phase of two orbit interference processing with the single view complex image of sentinel-1a. In the area analysis, phase estimation using adaptive regulation based on local smoothing (pearls) is used as the evaluation standard to compare this method with five phase unwrapping algorithms such as minimum cost flow. The results show that the square root error and the mean absolute error of this method are ± 0.0791rad, ± 0.0090rad, ± 2.3173rad, respectively. In the on-line analysis, for the absolute phase with only deformation phase, the deformation values of the strike line and tendency line of each unwrapping absolute phase face are extracted. According to the actual level observation value, the square root error, average absolute error and the maximum error of absolute value of the method in this paper are ± 0.17748cm, ± 0.14107cm, ± 0.40529cm respectively. Based on the phase reconstruction of pearls in surface analysis, it can be seen that the unwrapping method in this paper is better than other conventional unwrapping methods; similarly, in line analysis, the indexes based on level data are also the best.