| Three-dimensional (3D) pattern recognition technology with digital holography is a novel recognition technology, which has attracted extensive attentions. It is an organic combination of the booming digital technology with the conventional optical holography. Its high recognition sensitivity guarantees its potential applications in 3D pattern recognition and 3D measurement.This thesis mainly focuses on various recognition and invariance properties of 3D pattern recognition technology, as well as its comprehensive comparisons with traditional two-dimensional (2D) pattern recognition.In this thesis, a 3D opaque object has been established for the study of 3D pattern recognition and the propagation of its wavefront is simulated. We also discussed phase-shift technology which is used for wavefront sampling. According to the established 3D scene, we analyzed the convolution reconstruction method and Fourier reconstruction method. For the first time, we formularized the relationship between the 3D objects and their recognition results, which is independent of the reconstruction distance. According to our numerical experiments, the intrinsic higher correlation peak of 3D pattern recognition than that of 2D pattern recognition is revealed, whose difference is up to 105.Besides, the rotation-invariance, scaling-invariance and shift-invariance properties of 3D pattern recognition are labored in the thesis. The concept of "view angle" is used to realize rotation-invariance and measure the rotation angle of a 3D object. We first raised the scaling compensation method on holography plane to measure the planar scaling factor of a 3D object and analyzed the influence of the depth information of a 3D object on its recognition result.Finally, we extracted the wavefront distribution of a real optical field with phase-shift technology and reconstructed the field information with convolution reconstruction method and Fourier reconstruction method, which testified the validity of our theory.
|
PDF Full Text Request