Has anyone got experience in real time pattern recognition

asked 2019-10-09 02:19:05 -0500

genesis999 gravatar image

updated 2019-10-09 23:24:55 -0500

Hi all

I have a pattern below that is fed realtime to a window.

It contains only 4 colours

The idea is to split it up and smooth into sections. These shapes are labeled and segmented into horizontal rows. (blocky or pixelated divisions). ROunds up or down into these shapes. Any stepping blocks are considered reduction or expanding blocks.

We have X pixels height. updates from the right in realtime as time goes by. The idea is to create simple shapes that are stiched together and limited to a division of X pixels. Three shapes are squares, rectangles, expanding triangles and contracting triangles

If anyone has a solution or at least some means to start the ball rolling I will hire you.



edit retag flag offensive close merge delete


that is fed realtime to a window.

wait, do you have the actual data, that lead to the drawing ?

berak gravatar imageberak ( 2019-10-09 02:28:37 -0500 )edit

Yes. Data is here in this small app we are building

genesis999 gravatar imagegenesis999 ( 2019-10-09 07:06:03 -0500 )edit

I can send data as small bitmaps and you can stitch them together one after the other? and use pixel height divisions. Roundup Area over time.

genesis999 gravatar imagegenesis999 ( 2019-10-09 07:47:50 -0500 )edit

If you have the data leading to the drawing - maybe its easier to find the pattern there instead of using computer vision.

holger gravatar imageholger ( 2019-10-10 03:21:16 -0500 )edit

@genesis999, what is the problem, your program should solve ? is it visualization of the data ? are you trying to infer anything from that ?

berak gravatar imageberak ( 2019-10-10 03:28:01 -0500 )edit

yes to infer and block it up by horizontal divisions. Its kinda pixalated vertically and forced into a stepping kinda of shape detection. Other data is behind that infers what it is. Right now its just the means to convert into these blocks and steps. The reason for a visual approach is to check its accurate and its nice on the eyes too. its main goal are filter effects for audio usage. So if this pattern occurs filters this effect, if this occurs does another effect. (totally mad dubstep fat base effects) experimental stuff. so 2 stepping occurs goes into a fast filter start of shape to end. if its a step 2 down. does this swirling effect. if its 1 step 2 step on and off. Does some parabolic fast eq. mental stuff. Right now its just the shapes itself. visual just looks great.

genesis999 gravatar imagegenesis999 ( 2019-10-11 12:10:54 -0500 )edit

let me close this, it's a graphics design problem, not a computer-vision one , and entirely unrelated to opencv.

berak gravatar imageberak ( 2019-10-11 12:18:10 -0500 )edit

its an image = image detection = opencv is for image detection right?

genesis999 gravatar imagegenesis999 ( 2019-10-12 01:57:45 -0500 )edit