Home
--Research--Hybrid System
Hybrid System (Stereo Vision + Structured Light)
Related Articles
[1] Multicamera Phase Measuring Profilometry For Accurate Depth Measurement.
System Overview
Structured Light Illumination (SLI) is a method of range sensing involving the projection of a series of light patterns and allowing a camera to extract range data based on pattern distortions; whereas, Stereo Vision (SV) reveals depth through building correspondences between textured images recorded simultaneously by two or more cameras.
By employing multiple cameras, SLI can generate pattern phase/modulation and object texture data. To take advantages of SLI and SV, we developed a hybrid 3D reconstruction framework through phase-modulation-texture data fusion under temporal multiplexed illumination. The scheme is based on the error estimation of traditional SLI. Using only a small number of light patterns, our hybrid approach significantly reduces measurement errors versus traditional SLI methodologies while being insensitive to noise, conditions of texture and contrast, gamma distortion and projector flicker. Specifically, in our test, we achieved a 26 times reduction in error compared to traditional SLI. Currently, a technical article has been submitted.
The technology can be applied to various pattern strategies and applications especially for fusing 3D data from multiple views. The results show the advantages of this technology.
* Schematic diagram of a multi-camera and multi-projector 3D hybrid system.
Results and Analysis
* (a) A textured cartoon giraffe and a white space shuttle model. (b), (c) and (d) are zoomed-in and rotated views of (f), (g) and (h), respectively. (e) Depth rendering reconstruction by using stereo vision technique (with the graph-cut algorithm for 7 iterations). (f) Depth rendering reconstruction by using SLI technique (with 3 high frequency PMP patterns). (g) Depth rendering reconstruction by using SLI technique (with 30 high frequency PMP patterns). (h) Depth rendering reconstruction by using the hybrid approach (with 3 high frequency PMP patterns). (i), (j) and (k) are zoomed-in and rotated views of (f), (g) and (h).
* (a) A thick and shiny hair gorilla. (b) Depth rendering reconstruction by using stereo-vision technique (with the graph-cut algorithm for 9 iterations). (c) Depth rendering reconstruction by using conventional SLI technique (with 3 high frequency PMP patterns). (d) Depth rendering reconstruction by using the hybrid approach (with 3 high frequency PMP patterns).
Go to 3D Imaging Lab's home.
Copyright © 2009 3D Imaging Laboratory, University of Kentucky. All rights reserved.