Skip to main content
Open Access Publications from the University of California

UC Riverside

UC Riverside Electronic Theses and Dissertations bannerUC Riverside

Distributed Constrained Bayesian Optimization: Autonomous Camera Control


This dissertation describes methods to autonomously control an intelligent camera network with changeable pan, tilt, and zoom (PTZ) parameters for the purpose of obtaining high resolution facial imagery of randomly maneuvering targets. Every camera is treated as a self interested decision making agent that works in cooperation with the other agents in the network to attain a predefined system goal. The per camera per target image quality is designed and defined mathematically to formulate a distributed constrained optimization problem. Each camera is restricted to alter its own PTZ settings. All cameras use information broadcasted by neighboring cameras such that the PTZ parameters of every camera are optimized relative to the global objective. At certain times of opportunity, due to the configuration of the targets relative to the cameras, and the fact that each camera may track many targets, the camera network may be able to reconfigure itself to achieve a required target tracking specification for each target with remaining degrees-of-freedom. The remaining degrees-of-freedom can be used to obtain high resolution facial images from desirable viewing angles for certain targets. The challenge is to design algorithms that autonomously find these time instants, the appropriate imaging camera, and the appropriate parameter settings for all cameras to capitalize on these opportunities. The methodologies and solutions proposed herein involve a Bayesian formulation. The Bayesian formulation automatically trades off objective maximization versus the risk of losing target tracking performance. The dissertation describes a mathematical formulation of the visual sensing problem, design of functions that provide a measure of system performance, development of distributed methodologies that allows cameras to exchange information and asymptotically converge on optimal solutions, and incorporation of planning into the PTZ optimization methodology. The work herein presents theoretical solutions and analyses of results obtained on a simulated network of smart PTZ cameras.

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View