How to achieve real-time human silhouette and movement detection?

I started with background subtraction between a static background and another image with a foreground object(hand) that were captured from webcam. I couldn't get good results. I have attached the code i used and images below.

[Walter Roberson]

You do not appear to be accounting for illumination differences. Possibly that would help.

Answer 1

[Image Analyst]

You can avoid the loop by saying

binaryImage = Out > 3; % Or some number that works.

Then use bwareaopen() or bwareafilt() to get rid of small blobs.

[Avinash S]

i cant get u. Can u please tell me where to modify the code?. I'm new to this stuff

[Image Analyst]

Can you make it easier to help you? Can you attach your separate images?

[Avinash S]

Pic: Background

Pic: Current frame

Answer 2

[Image Analyst]

See my attached code.

If you want a smoother background, like to not have the movement of the drapes shadow be connected to the arm, you can use activecontour(). I've also attached a demo for that so you should be off to a good start.

Note: This may not be the best approach - the foreground detector in the Computer Vision System Toolbox may work better.

[Avinash S]

thank u so much for your great effort. But I couldnt get best results using real time input stream from webcam. Now I came across this paper based on fuzzy logic which looks very good in extracting silhouette: _ Interval Type-2 Fuzzy Logic System for Human Silhouette Extraction in Dynamic Environments_ https://www.kau.edu.sa/Files/611/Researches/63095_34134.pdf

Is there an implementation for this? or can u help me implement this

[Image Analyst]

Sorry, I can't. Besides, I don't even have the Fuzzy Logic Toolbox.

[shine day]

hi! Image Analyst. could you give me your email? I would like to ask you about image background subtraction. Best regards!

Answer 3

[Dima Lisin]

You can use vision.ForegroundDetector in the Computer Vision System Toolbox.

[Avinash S]

can you give me some example?