Roland from South Africa  [107 posts]
10 year

Is there a simple way to integrate the result of a test?

When I move my camera to a vantage position, I want to perform a test and move away as soon as possible, but only after the confidence has reached a certain level, negative or positive.  I only see tools to integrate frames.

I see the 'shape matching' has a confidence filter, but does this apply to blob counting too?

Steven Gentner from United States  [1446 posts] 10 year

I'm not sure I understand what you are asking for? Perhaps you are looking for the Filter Variables module that will average variables over time to reduce possible noise that may be inherent in the results.

If not, can you better explain what the problem is?

There is no confidence in blob counting, a group of pixels is either a blob or not, there is no confidence needed in a yes/no situation. The shape match has a confidence since shapes can seem like other shapes but not exactly.

Roland from South Africa  [4 posts] 10 year

Thinking aloud here;

I have a set of filters set up at 'position 12' in my robot sequence. At some point the camera swings to position 12 and a set of filters is enabled, maybe;
- threshold
- blob filter
- setvariable ("here"), COG

What I don't know is how long to wait at this position before moving on.
Is each filter fully processed before the next, or is each step processed irrespective, based on a time slot? Could I get a 'wrong' COG on the first pass? Then, if so, I can't move simply on the basis of a result. It could be a false-positive.

Otherwise, if I use a timer, I may be wasting a lot of time, or too short. How do I know that all the filters have been processed to the degree that the result is 'good'. I'm not talking about an unstable image. The image is stable, but the white balance may still be stabilising. I just don't know when to move on. I looked at the 'filter variable' function, and while it does averaging, there is no flag to indicate anything about it.

Ideally?, in the 'filter variable' window, there would be a % setting, and a 'done' flag. So lets say I set 'here' as the variable, and 20 passes and 80%
It would then load 'here' on successive passes and based on continous results, if the value was consistent for 80% of 20 passes, the 'done flag would be set. In my program I would have an 'if 'done' = 1' then move on function. I suppose I can do the math myself.

So, this is not a feature request, I'm just trying to figure out how to determine when a result is a reasonable deduction.

On the other hand, I suppose if I took the mean of 10 passes, and just counted 10 off IMAGE_PROCESSED, would that be as good as it gets?
The camera runs mostly at about 8fps.

Steven Gentner from United States  [1446 posts] 10 year

What we normally do in these kinds of circumstances is to wait until the image is stable ... i.e. you see more than 1 frame that is very similar to the previous one. Eval that frame and then move based on that.

The movement module can be used to determine how much of an image is changing based on the variables it creates.

In terms of your thoughts about the 20 passes and 80% consistent, perhaps the closest measure is the delta selection in that module. That will give you a difference between the current value and the mean of the last 20. That's basically variance which means that the current value is inline with the average of all past values and a good indication that things are stable. Have you tried that option?


This forum thread has been closed due to inactivity (more than 4 months) or number of replies (more than 50 messages). Please start a New Post and enter a new forum thread with the appropriate title.

 New Post   Forum Index