.png){kind=icon}
.png){kind=icon}
Return
Deep-Learning-Based Single-Shot Fringe Projection Profilometry Using Spatial Composite Pattern
.png){kind=icon}
PRE
AI
DOI：10.1109/TIM.2024.3420365.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Single-shot fringe projection profilometry (FPP) is crucial for real-time or dynamic 3-D measurement scenarios. In this regard, we propose a spatial composite FPP (SCFPP), which creatively fuses the spatial feature in different frequency fringe patterns. Inspired by image inpainting techniques in computer vision, SCFPP employs a novel adaptive space-division encoding strategy (ASES) to divide fringe patterns into several basic blocks and reaggregate partial basic blocks in the spatial domain to form a spatial composite fringe pattern. In the demodulation stage, we develop a deep-learning network model, the phase inpainting network (PIN), to restore the spatial composite fringe pattern to complete phase information. The absolute phase map is then obtained with high accuracy by applying the conventional three-frequency heterodyne phase unwrapping (CTHPU) algorithm with three known wrapped phase maps. To the best of the authors' knowledge, SCFPP is the first successful use of spatial multiplexing strategy in FPP. In addition, SCFPP can be seen as an application of image inpainting techniques in the fringe projection field, introducing new ideas for applying computer vision methods in fringe analysis.
Keywords:
Image color analysis
Accuracy
Phase measurement
Frequency-domain analysis
Extraterrestrial measurements
Demodulation
3-D measurement
deep learning
fringe projection
image inpainting
single shot
single shot
Journal
IF:
5.9
Papers:
1.9W
Citations:
5.8W