Self-localizing Smart Cameras and Their Applications

Publication Year:
2012
Usage 43
Downloads 33
Abstract Views 10
Repository URL:
https://repository.upenn.edu/edissertations/579
Author(s):
Shirmohammadi, Babak
Tags:
Distributed tracking; Multi-Camera Tracking; Self-localizing Cameras; Smart Cameras; Computer Sciences
thesis / dissertation description
As the prices of cameras and computing elements continue to fall, ithas become increasingly attractive to consider the deployment ofsmart camera networks. These networks would be composed of small,networked computers equipped with inexpensive image sensors. Suchnetworks could be employed in a wide range of applications includingsurveillance, robotics and 3D scene reconstruction.One critical problem that must be addressed before such systems canbe deployed effectively is the issue of localization. That is, inorder to take full advantage of the images gathered from multiplevantage points it is helpful to know how the cameras in the sceneare positioned and oriented with respect to each other. To addressthe localization problem we have proposed a novel approach tolocalizing networks of embedded cameras and sensors. In this schemethe cameras and the nodes are equipped with controllable lightsources (either visible or infrared) which are used forsignaling. Each camera node can then automatically determine thebearing to all the nodes that are visible from its vantage point. Byfusing these measurements with the measurements obtained fromonboard accelerometers, the camera nodes are able to determine therelative positions and orientations of other nodes in the network.This localization technology can serve as a basic capability onwhich higher level applications can be built. The method could beused to automatically survey the locations of sensors of interest,to implement distributed surveillance systems or to analyze thestructure of a scene based on the images obtained from multipleregistered vantage points. It also provides a mechanism forintegrating the imagery obtained from the cameras with themeasurements obtained from distributed sensors.We have successfully used our custom made self localizing smartcamera networks to implement a novel decentralized target trackingalgorithm, create an ad-hoc range finder and localize the componentsof a self assembling modular robot.