Page 1 of 1

blob detection

Posted: Thu Nov 08, 2018 2:39 pm
by venkateshn
in blob detection i am able to identify the object, i wan to find out that how much time it is taking to detect blob object (single or multiple blob object).

Re: blob detection

Posted: Thu Nov 08, 2018 3:19 pm
by kwagyeman
Hi, this is just the inverse of the FPS. So, if you're running at 30 FPS then it's 33ms.

Re: blob detection

Posted: Thu Nov 08, 2018 11:28 pm
by venkateshn
What is the SI unit of fps?? is it hertz in this case

Re: blob detection

Posted: Sat Nov 10, 2018 10:26 pm
by kwagyeman
Yes.

Re: blob detection

Posted: Mon Nov 12, 2018 12:34 am
by venkateshn
how to measure latency and accuracy in blob detection experiment

Re: blob detection

Posted: Mon Nov 12, 2018 12:10 pm
by kwagyeman
? You have to write some more detail on what you want.

Re: blob detection

Posted: Mon Nov 12, 2018 12:21 pm
by venkateshn
I am doing marker based detection with normal color led like rgb colors.. I am using blob detection single color script for tracking leds.. Since it is finding blob.cx() and blob.cy() I want to find out how accuratey it is tracking and finding blob.cx and blob.cy

Re: blob detection

Posted: Mon Nov 12, 2018 12:36 pm
by kwagyeman
I see. We have no metric for this.

Um, cx and cy are just the sum of x/y pixel positions in the tracked color list divided by the number of pixels tracked. Google centroid.

If you want a tracking confidence... then just use the density() method to determine if the object is solid or not. density is the number of pixels divided by the size of the bounding box. When low it means the lock is bad. When high it means the lock is better (i.e. you are tracking all pixels well).

Re: blob detection

Posted: Tue Nov 13, 2018 4:24 am
by venkateshn
in optical flow you are using phase correlation concept, what is the concept and logic behind blob detection
how is img.find_blob working internally, is it undergoing any transformation

Re: blob detection

Posted: Tue Nov 13, 2018 11:59 am
by kwagyeman
Um, the algorithm for that is somewhat custom. Basically you scan the image looking for pixels that match the thresholds... once you find one you then recursively connect all pixels that share the same thresholds and keep track of what pixels you've visited. Then you continue scanning after enumerating all pixels in an object. Read the code if you'd like to know more: https://github.com/openmv/openmv/blob/m ... img/blob.c

Re: blob detection

Posted: Thu Nov 15, 2018 1:33 pm
by venkateshn
In blob detection you have defined threshold of six tuples of LAB color space, when tracking ,Suppose i am tracking red color does threshold of all six tuple values defined for red should match with target red color or is any one of the threshold value is matching .

Re: blob detection

Posted: Thu Nov 15, 2018 6:59 pm
by kwagyeman
Yes