Detecting video feed + taking high quality pictures

Alain

Is there a way to have a camera handle the detection (at high frame rate using H.264) and once the detected object cross a specific point, a picture is taken and the camera returns to detecting objects.

The tricky part here is I want the picture to be taken using the raw image, not a compressed frame from the video feed.

lovro

Hi,
yes — you can run detection on the video stream, then check if the centroid (i.e. the center point of the bounding box) crosses your target line or is in the center of the frame. When that happens, trigger a high-res capture from the raw (still) image stream and save it using cv2.imwrite().

Alain

lovro Thanks! I looked at code sample and I wasnt clear if using the compressed video from h264 for detection would allow access to the raw stream.

lovro

Hi,

If you want to access the raw stream, take a look at the ColorCamera node documentation:

ColorCamera Node

It explains the different output types (raw, isp, video, still) and how to use them to get uncompressed or full-resolution frames from the camera.