odissey1
November 15, 2006, 18:53:25
Hi,
I have a camera DFK 41AF02 which is mounted on the helicopter to survey area and return images at a rate of 1 image/sec. To reduce the CPU load I interrupt video stream using
ICImagingControl1.LiveCapturePause:=True (or False);
If in the camera properties I enable AutoWhiteBalance, AutoExposure, AutGain etc, then It takes camera about 10 iterations to adjust these parameters (which is equivalent to 10 second) which is too slow for that application. For example, moving from sandy desert landscape (light background) to a forest (dark) may occur in a matter of a seconds, and camera does not adjust itself quickly enough.
Q1. Where processing of the image and subsequent adjustment of the camera settings is actually performed: onboard (in the camera itself by means of FPGA/microcontroller) or inside the PC by means of WDM driver?
Q2. Why camera settings adjustments are so slow? Could iterations converge more quickly? To be more precise: I would expect that a single image should be sufficient to estimate an exposure for the next one. Is it worth trying to implement such an algorithm inside the PC (just guessing)?
Sincerely,
Boris
I have a camera DFK 41AF02 which is mounted on the helicopter to survey area and return images at a rate of 1 image/sec. To reduce the CPU load I interrupt video stream using
ICImagingControl1.LiveCapturePause:=True (or False);
If in the camera properties I enable AutoWhiteBalance, AutoExposure, AutGain etc, then It takes camera about 10 iterations to adjust these parameters (which is equivalent to 10 second) which is too slow for that application. For example, moving from sandy desert landscape (light background) to a forest (dark) may occur in a matter of a seconds, and camera does not adjust itself quickly enough.
Q1. Where processing of the image and subsequent adjustment of the camera settings is actually performed: onboard (in the camera itself by means of FPGA/microcontroller) or inside the PC by means of WDM driver?
Q2. Why camera settings adjustments are so slow? Could iterations converge more quickly? To be more precise: I would expect that a single image should be sufficient to estimate an exposure for the next one. Is it worth trying to implement such an algorithm inside the PC (just guessing)?
Sincerely,
Boris