halley's profile - activity

What's the best way to proceed, to add support for a new camera device?

The Raspberry Pi has a specialized camera interface, but it does not follow/support Linux standard kernel mechanisms to access the data.

The Raspberry Pi Foundation has written only some userland C demo code that directly accesses the hardware and retrieves stills or h254 video. Since these are demo apps, the only way they output data is back to the slow SD disk. There is other overhead in starting/stopping the demo app for each frame, as well.

What a lot of robot hobbyists are looking for is the "right way" to add a C module/plugin/API that allows all OpenCV code (including python bindings, etc.) to get at those images with the minimum setup and memcpy overhead.

I know not much about OpenCV's data acquisition architecture, but am interested in diving in.