Real-Time 3D-Reconstruction With GigE Vision
The 3-D Imaging Lab at the University of Kentucky’s Center for Visualization and Virtual Environments has extensive experience in the area of 3-D and is credited with developing 3-D systems in the fields of machine vision, structured light system and real-time 3-D scanners, biometrics, next generation multimedia, and more.
Real-time 3-D Structured Light System
In 2006, the University of Kentucky received a grant from the US Department of Homeland and Security to develop a method of range sensing for 3-D surveillance cameras.
The resulting product is a high-resolution, real-time 3-D system that employs Structured Light Illumination (SLI) as well as high-speed reconstruction methods to form 3-D images. Structured Light Illumination is the process of projecting a series of striped line patterns so that, viewed from a certain angle, a digital camera can reconstruct a 3-D model of an object’s surface based on the line distortions that occur on this object.
The system consists of a high speed Prosilica VGA resolution camera with Gigabit Ethernet output and a projector. The camera was chosen for its fast frame rates (up to 200 fps at full 640 x 480 resolution) and plug-and-play attributes. A monochrome camera is preferred to a color model for this application as the use of color would impact on the system’s performance since some colors would not be able to reflect on others (e.g. red on green). The projector is composed of a Texas Instrument Discovery 1100 board with ALP-1 controller and LED-OM with 225 ANSI lumens.
The camera is fitted with a Tamron 16mm lens and mounted below the projector. The camera and projector are separated by 10° for optimum area scanning results, and synchronized to operate at 120 frames per second at 640 x 480 resolution using an external triggering circuit. The projected light patterns are loaded directly into the ALP board’s memory.
The system is connected to a Dell Optiplex 960 with an Intel Core 2 Duo Quad Q9650 processor running at 3.0 GHz via CAT-5e cable.
The system is fully scalable and the camera optics, projector light brightness and viewing angle can be easily changed to adapt to a larger scale application.
Real-time 3-D point clouds reconstruction at 120fps
SLI Pattern Strategies
The system uses a range of 3 to 6 different alternating linear patterns depending on the conditions (i.e. moving or stationary object). Each pattern contains 16 pairs of black and white stripes (periods) that are projected on to the target(s) in a pre-determined sequence of flashes at 120fps.
The University of Kentucky developed three new pattern strategies to ensure the quality and accuracy of the 3-D reconstruction: A Period Coded Phase Measuring and a Dual Frequency Phase Multiplexing pattern that unwrap the high frequency phase without projecting more patterns or using complicated spatial unwrapping methods, and an Edge Pattern Strategy that increases the signal to noise ratio of the system.
High Speed Reconstruction
While other systems generate 3-D point clouds on a pixel by pixel basis using an inverted 4x4 matrix, the 3-D Imaging Lab has developed new look-up tables that replace the repetitive and constraining matrix process and allow the system to decode the phase video at the full 640 x 480 resolution and 120 frames per second rate. No recording or post-processing is used.
The 3-D point reconstruction technique is applicable to all triangulation-based 3-D techniques including SLI, stereo vision, laser scan, etc.
The 3-D reconstruction software was developed in Microsoft Visual Studio 2005 with managed C++ using the Prosilica Software Development Kit.
The University of Kentucky’s real-time 3-D Structured Light System is fully scalable and can be adapted to perform in applications in the areas of human computer interfacing, biometrics and security, motion scanning and tracking, hand gesture, facial recognition, fast 3-D modeling, and next generation multimedia.
The system is available for licensing for use in commercial products.