如何在 OpenCV/Python/OSX 中从 PointGrey USB 相机捕获帧？

Question

PointGrey is a leading manufacturer of machine vision cameras, but unfortunately their support for Mac OS is very limited. A www search led me to guess that I need to install libusb and libdc1394 in order to recognize the camera, which i did using brew. This did not work.

I don't understand exactly how libusb and libdc1394 libraries work under the hood, other than that they handle the hand-shaking with the camera and data transfer via the USB bus. OpenCV usually makes it incredibly easy to open a camera and start processing frames, but unfortunately when it doesn't work it's not clear how to debug. I found python wrappers for libusb and libdc1394 and included them in my code, which resulted in no errors, but no luck grabbing frames either.

If you don't have an exact answer, suggest strategies for solving this problem ie how should I systematically approach this, diagnosing all the possible failure modes. Is there a way that I can see more of what's going on when OpenCV tries to detect and read from a camera?

My python/opencv code works well with a simple capture = cv2.VideoCapture(0) but doesn't work with capture = cv2.VideoCapture(1) , giving output as follows:

Warning, camera failed to properly initialize!
Cleaned up camera.

Typically cv2.VideoCapture(0) will give the built-in camera on my macbook, and from what I understand cv2.VideoCapture(1) will give the next available camera (ie plugged in through USB).

I know the camera works well on a windows machine (in Windows Movie Maker). Do I need to do something further under the hood to get python and opencv to recognize the camera under OS X?

Answer 1

Many years later, we can answer this question. FLIR, which acquired PointGrey, has released Spinnaker, a cross-platform SDK with a UI and a set of pre-compiled examples for accessing the camera: https://www.flir.com/support-center/iis/machine-vision/downloads/spinnaker-sdk-and-firmware-download/