|
|||
|
|||
|
Managing the Information Flow in a Network of Visual Sensors
|
 |
| Randomized Invariant Features for Shape Classification funded by NSF, Computer Vision Program We are studying the behavior of image classification systems when only a limited number of training samples are available. We have given qualitative and quantitative evidence of small-sample phenomena in the context of invariant operators and of the fusion of multiple classifiers. Related papers:
|
 |
| Exploring the World with a Ray of Light: An Environmental Sensor for the Blind funded by NSF - Sensors and Sensor Networks Program We are building the "virtual white cane", a laser-based mobility tool for the blind. It allows the user to measure distances to surfaces, and to detect environment features that are important for safe deambulation, such as steps and drop-offs. We have developed a prototype hand-held system, and designed tracking algorithms for feature detection. Here is some more information about the project. This project was covered by UCSC's Currents, the San Francisco Chronicle, the Santa Cruz Sentinel, Laser Focus World, Santa Cruz's Good Times, and Science Today. Related papers: |
|
| Wayfinding for the Blind & Visually Impaired Using Passive Environmental Labels funded by NIH - National Eye Institute with J. Coughlan, SKERI We are developing a wayfinding system for the blind and the visually impaired, which use a regular camera cellphone to detect special color targets in the scene. The targets, which contain contain semantic and spatial information, are very economical to make. The user scans the environment with the cell phone, which detects all landmark symbols in its line of sight up to distances of 10 meters. Once a landmark is detected, the size and shape of the landmark symbol as imaged by the camera determines the approximate distance of the user from the landmark. Related papers: |
 |
|
||||||