Three-Dimensional Reconstruction System based on a Segmentation Algorithms and a Modified Fourier Transform Profilometry

摘要

In this work, a modified Fourier Transform Profilometry method, together with the region post-processing (where the object to digitize is located) is used as a 3D reconstruction system for solid objects. By projecting a sinusoidal fringe pattern with a known spatial frequency on the object, a vision system is capable of infer the object's depth information. First, the fringe pattern is projected on a reference plane, later the object to digitize is placed in front of the reference plane and the same fringe pattern is projected. Both images, the reference and the object, are acquired and the proposed method is applied. One of the main difficulties is the high frequency content in the unwrapping process, and this effect can be minimized by knowing the spatial frequency and its multiples. It can be observed that the fringe pattern is distorted due to the object's irregular shape in comparison with the reference plane. At this time, the modified Fourier Transform Profilometry Method (FTP) is applied where the resulting phase difference of the acquired and processed images has the information of the object's depth. To improve the results, the object's contour is taken into account to process only the object area and avoid the errors produced by shadows. Some tests with computer generated and real objects with different geometries are carried out to verify the proposed methodology, considering the factors that mostly affect the method.