I have compiled and run the hd-camera sample application supplied by Intel at https://github.com/intel-aero/sample-apps/tree/master/hd-camera sample-apps/hd-camera at master · intel-aero/sample-apps · GitHub to test the RGB HD camera (OV8858) on the Intel Aero (/dev/v4l/by-path/pci-0000:00:03.0-video-index2) but the images are very underexposed so the images are quite dark. RGB pixel values are in the 0-30 range out of 255 in a normally lit room. Shining a flashlight directly at the camera can yield an RGB pixel value of (255,255,255), so the range of pixel values are not being saturated. It looks very much like the default exposure settings are quite wrong, and standard V4L2 mechanisms for setting the exposure manually do not appear to have any effect.
Does Intel have any examples akin to hd-camera that actually provide normally exposed camera images from the HD camera? The RealSense camera works very well and is properly exposed, as does the optical flow camera. We have been trying for some time to get the HD camera to work properly, and while we can get images from the camera, both with the examples supplied by Intel (hd-camera, capturev4l2) and with our own V4L2-based code, the exposure always appears to be far too low in all cases. I have pasted an example below. I included the overhead light so you can compare pixel values in the full spectrum of light to dark. Note that the room is well lit and the camera does not appear to adjust the exposure automatically (without a light in the image it's just dark). Do you not observe the same behavior from the hd-camera sample or capturev4l2 sample? The image was taken from our own software on the Intel Aero but the sample applications produce very similar results.
Help would be greatly appreciated.
Thank you for your Interest in Intel products.
In order to help you further, I need you to answer a few questions first:
What values did you choose for aperture and shutter speed?
With the camera's automated settings, please take a picture of a white, black and 50% gray cardboard and send all three of them back to me to take a look.
With the exposure of the 50% gray please check where the information goes on the histogram and if the image is underexposed how much do you have to compensate to bring it to zero?
It seems that there are some limitations with the down facing camera.The engineering team is looking into it. We will let the community know as soon as we have an update.
The RGB camera I'm referring to is on a swivel mount and faces forward and down, as opposed to the greyscale optical flow camera which faces directly downward. I'm assuming we are talking about the same RGB camera but thought I should clarify just in case.
I have tried adjusting brightness, hue, sharpness, color effect, backlight compensation, contrast, saturation, gamma, white balance, gain, pan, roll, exposure, focus, tilt, zoom and iris settings in our own V4L2 application and none of them appear to do anything. The same application talking to the RealSense camera can change some of those settings, so I don't think it's a bug in the code. However, since you appear to be confident that the camera aperature and shutter speed can be set, do you have some sample code that sets these parameters of the camera? Also, I should note that the camera does not appear to automatically adjust any settings despite high and low-level light conditions.
Here is a picture of a white sheet of paper:
Here is a picture of a black object:
Here is a picture of a grey object (not guaranteed to be 50% grey I'm afraid):
Finally, here is a picture in normal lighting of my hand:
All camera settings are defaults. However, I can't seem to change any of the camera settings mentioned above via V4L2 on this camera anyways. Note that these images may show artifacts from JPEG compression because I compressed them but other than that they are very representative. Just as a point of comparison, here is a picture taken using the RealSense camera of my hand using the same application, same lighting conditions, same JPEG compression and default camera settings:
Thank you for your answer. It seems that for this camera also is a known limitation and the engineering team is currently looking into it. As soon as we have additional information, we will update the thread.