Subscriber Discussion

Any IP Cameras Adjust Shutter Speed/Gain Based On Motion Detection?

If not, why not? Many cameras adapt to motion by increasing the frame rate, why not change the shutter speed/gain?

1. Enable camera to capture scene without bluring during periods of quick motion. (Person running towards a entrance)

2. Enable camera to capture scence with enhanced detail/reduced artifact during more static scenes. (Same person pausing at a an entrance)

3. Reduce nighttime noise/bandwith by using less gain unless "real" motion is detected.

4. Make motion detection at night more usable because of less false positives.

Is the problem that you can't change the exposure settings of camera quickly enough, i.e. between frames? Or that the transition frame(s) would be somehow artifacted and unusable? Maybe the switching back and forth could be too quick, and would certainly bloat the H.264 stream during those transitions. Maybe a dual MJPEG stream and H.264 stream with different exposure settings for each? Does that exist?

Good idea / question. I think it's interesting and doable but I do not know of anyone who offers this. Let's see if anyone has any suggestions or knows of someone doing this.

The only real answer to low light is illumination! Digital manipulation will never solve what is an unmistakable need for cameras, light!

But light is not always available and/or people do not want to pay for lighting.

In those common scenarios, people look for tricks / optimizations to compensate.

Digital manipulation will never solve what is an unmistakable need for cameras, light!

Paolo, this is not digital manipulation. This is just automatically slowing the shutter speed in times of little motion, thereby allowing more light in to capture more detail, but then as motion is detected scale the shutter speed to avoid blur.

Rukmini, the metapoint is that it is 'manipulation' of some sort. But lots of people don't have enough light, so this one would a useful 'manipulation' / trick for them.

One downside I can think of it support calls - "Hey when people move the video gets really dark, what the hell is happening!" On the other hand, they likely already get support calls "Hey when people move the person looks like a ghost, what the hell is happening!"

Ok, its manipulation of digital control. Though I think there is an important distinction to be made between DSP manipulation, (which is essentially taking sampled numeric values and performing various mathematical transformations upon them), and modifying things that are under digital control, (like AutoGain), that affect the image before A/D even takes place.

Paolo's statement of 'digital manipulation will never solve what is a... need for cameras, light!', leads me to believe he thinks it is a DSP 'trick'. But since in actuality, this IS providing more light, he should be satisfied.

One downside I can think of it support calls - "Hey when people move the video gets really dark, what the hell is happening!

In a perfect world, the overall brightness would not change when people move, because the gain would be adjusted at the same time as the shutter speed to compensate. Of course the character of the stream might fluctuate quite dramatically though, leading to similar support issues...

I know some cameras which can use double profiles, so 2 different shutters / gain/ Iris settings / contrast/ brightness/Sensor Exposure./even Motion .etc (One auto profile at daylight then manual at night) but you can't change Fps or codec the same way. THat 's the type of camera which can adapt to extreme different brightness.

I know some cameras doing Adaptive streaming: so transmiiting only I-Frames when nothing happens with motion detection (in fact dropping frames kind of SVC) , and then transmit regular stream speed when alarm accur. (but you need some Sdk integration on the Vms to take advantage of it)

Then I know that some manufacturers using some ..sensor/Dsps, can handle large dynamic or small dynamic ranges, handle 60 fps or 30... but these cameras has to reboot lighthly the stream engine when you want to change the mode (by CGI command for example) .. and you lose stream during 6-9 seconds.....

so Technical issues at the time being to get what you expect, but you are right will propobably come, because it's a common need to all of use

Marc, do you know how often an AGC will 'evaluate and adjust' the analog gain setting of a sensor? Is it feasible to do every frame? Whatever the interval is, do you see any technical limitation to changing both gain and shutter speed on each frame?

If you could then I suppose you could multi-stream a high gain, fast shutter stream AND a low gain slow-shutter one, by alternating the settings on successive frames, therby giving you the best (and the worst) of both worlds. You might also be able to multi-stream B/W and color at the same time if you can switch them quickly enough, and as long as there wasn't any noticeable effect from hysteresis.

At the time being what I know is that many Leaders would simply love to have a simple double profile and haven't. So I presume the CPU used by a one frame analysis, would be dramatically high

I think you can figure it with the current Wdr : software version,or Dsp , with 2 or 4 frames

When using 4 frames thanks to Dsp, you can't no longer play with shutter speed or Gain, neither get fast fps, so again a trade off among multiple parameters

One common mistake by SI is to activate too many features in the camera: 2 or 3 streams, 2 motion areas, Wdr, + Dnr + ..... and discoverfianly at night that the camera is freezing or losing frames. (even on high end cameras) I try to fight to get a CPU level in the cameras and asked for it to several manufacturers to also decrease technical support.

Bosch did it in one model I recently received. You get a "Linux" CPU usage, in real time. First one.

I'm sure thanks to Moors law we can get soon much more CPU power so much more intelligence within most cameras