Geometric constraint phase unwrapping method based on self-supervised deep learning

The invention relates to a geometric constraint phase unwrapping method based on self-supervised deep learning, and belongs to the technical field of image processing. The method comprises the following steps: S1, acquiring a fringe picture of an original to-be-measured object through a three-dimensional measurement system, calculating to obtain a wrapped phase diagram and a background light intensity image, and obtaining calibration parameters of a projector and a camera in the system through calibration; S2, converting the wrapped phase diagram and the background light intensity diagram in the S1 into an image of a stripe-level required by phase unwrapping through a convolutional neural network; and S3, performing phase depth mapping and corresponding system calibration parameter calculation on the stripe-level image in the S2 to obtain accurate three-dimensional information. The problems of low generalization ability and strong data dependence existing in phase unwrapping based on supervised learning can b