Research Statement
My long term research goal is visual understanding through extended robotic interactions with the world. I am interested in processing EO/IR video from an actively moving image sensor over long periods. My goal is to generate useful visual measurements with predicted error characteristics, to enable autonomous intelligence, surveillance and reconnaissance tasks in robotics. I seek to create a robust robotic vision system that can solve probabilistic visual inference and generalized correspondence problems in unconstrained outdoor environments. Broad themes of my work have been computer vision to enable outdoor robotic autonomy, visual collision detection, image segmentation using perceptual organization, inertial measurement aiding for vision, visual navigation and real time embedded video processing.
Projects
My current work is focused on computer vision for unmanned air vehicles (UAVs):
VAMAV: Visual obstacle detection and avoidance for a micro air vehicle. Monocular image segmentation and area moments for time to collision estimates.
VISTA: Visual Threat Awareness for an Unmanned Air Vehicle. Stereo based obstacle detection for small UAVs.
ImageNav: Image aided navigation for an unmanned air vehicle in GPS denied areas.
EVASE: Supervised learning and 2D pattern classification in noisy video.
K-cut Image Segmentation: Perceptual organization using multiway graph cuts.
Constrained Fuzzy Clustering: Constrained fuzzy clustering for target detection in particle filters.
Completed projects at the Carnegie Mellon University Robotics Institute (CMU-RI):
Reconfigurable Vision Machine: Research staff for modular architecture and software tools for high performance machine vision.
Micro Vision Engine: Consultant on low noise analog NTSC layout for a vision computer suitable for a small unmanned vehicle.
TUGV: Research assistant on high speed cross country navigation using stereo vision for the Transitional Unmanned Ground Vehicle.
Virtualized Reality: Research assistant on hardware and software based camera synchronization for multibaseline 3D reconstruction.
Automated Timber Inventory: Precise texture segmentation for tree diameter estimation.
Biography
Jeffrey Byrne is a senior research engineer at Scientific Systems Company Inc. (SSCI) in Woburn MA, working in the areas of Computer Vision and Robotics. He has a BS (1996) and MS (1997) in Electrical and Computer Engineering (ECE) from Carnegie Mellon University, and he is also affiliated with the GRASP Lab at the University of Pennsylvania under an NDSEG fellowship . Prior to joining SSCI, Mr. Byrne worked as a consultant to Carnegie Mellon University on the Micro Vision Engine hardware platform for micro air vehicles (MAV), and as a research staff member in the CMU Robotics Institute working on image segmentation and classification, high-speed vision based inspection, real time vision for helicopter perception and stereo vision based autonomous mobile robot navigation. He also worked as a engineering manager in an embedded hardware start-up company in Silicon Valley. He is a US Citizen and has a Secret level clearance.