Multicamera Array Matt Casella Elizabeth Dinella Killian Coddington
Multicamera Array Matt Casella, Elizabeth Dinella, Killian Coddington, Nate Bellavia
Origins of project • Motivated by work done by Stanford Ph. D students • RIT Prototype developed during 20122013 academic year by students in the Freshman Imaging Project class
Synthetic Aperture Imaging Reference: Marc Levoy
Basic Concept Cameras Occluder Object
What is a Multi-Camera Array?
Prototype Architecture Mac Laptop Camera Image capture x 6 Image compression and resizing Computer Switch Data routing x 6 User Interface Arduino Trigger cameras Select operating mode, initiate operation, render imagery Processing, control, storage
Imagine. RIT Prototype
Summer Objectives • Increase frame rate of prototype by improving image processing • Develop and validate a predictive model • to facilitate analysis of alternate configurations Conduct cost/performance trade-off study
Bottlenecks • Matlab processing limitations o Handling of image display • Non-optimal data flow o Picture transfer from camera to software
Components of the software • Calibration • Synthetic aperture through image shifting • GUI
Progress to date Matlab to C++ translation complete • Result: 400% faster processing o 16 frames per second • Limitations: capture rate of cameras
Next Steps • remove remaining bottlenecks o File stream • Software optimization
Summer Objectives • Increase frame rate of prototype by improving image processing • Develop and validate a predictive • model to facilitate analysis of alternate configurations Conduct cost/performance trade-off study
Predictive Model
Static Validation
Dynamic Validation
Model Applications • Alternate Configurations o Different cameras o Number of cameras o Placement of cameras
Cost/Performance Tradeoffs • Point Grey cameras used in prototype cost ~$650 each • Other potential cameras available for a cost of ~$25 each • Planning to model and test system using • Raspberry Pi cameras to determine impact on image quality and frame rate Work currently in progress
Summary • Demonstrated significantly improved performance by translating processing routines • Validated predictive model to allow rapid assessment of alternate architectures • Anticipate completing cost/performance trade offs by end of summer
Acknowledgements Dan Smialek Joe Pow RIT Carlson Center Facilities Manager RIT CIS Associate Director Philip Salvaggio Dr. María RIT Imaging Science Helguera Ph. D Student RIT Associate Professor Stefi Baum Billy Frey Director of CFC Center for Imaging Science undergraduate student Dr. Roger Easton RIT Professor Jason Faulring RIT Systems Integration Engineer Lindsey Schwartz Imaging Science undergraduate student Bob Callens Intern Program Coordinator Dr. Carl Salvaggio RIT Professor
- Slides: 21