This preview shows pages 1–2. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.View Full Document
Unformatted text preview: 1 Time to Collision Warning Chip: Background: Under certain circumstances, it is possible to estimate the time to collision (TTC) from a time varying image. It would be useful to encapsulate an algorithm for doing this into a cheap camera, ideally with computation done on the chip that senses the image. That way, there is no need to move large volumes of image data from an image sensor to a processing chip. Such a chip would perform an extreme bandwidth compression: it has high bandwidth in (image sequence) and low bandwidth out (time to collision). Such a device could be outfitted with a cheap plastic lens and used as a warning system, aimed, for example, out the rear of car into each of the two blind spots that are not easily visible in the drivers mirrors. The key to recovering the time to collision is the realization that there are constraints between the brightness gradient (spatial derivatives of brightness) and the time derivative of brightness at a point in the image. These depend on motion field, and the motion field in turn depends on the rigid body motion between the camera and the object(s) being viewed. Focus of Expansion: A related chip is the focus of expansion (FOE) chip built by Ig McQuirk. This is an analog VLSI chip that finds the focus of expansion. His chip uses spatial and time derivatives of brightness to locate the point in the image towards which motion is taking place. Again, this is a high band width in, low band width out application. The difference between the two projects (FOE versus TTC) is that in the FOE chip, the intent is to be insensitive to the magnitude of the ve- locity, and to distances to points in the scene, while in the TTC chip the opposite is the case: the intent is to be insensitive to the direction of mo- tion instead,...
View Full Document
- Spring '04