may i know how we want to measure the depth for car tire thread?

currently im doing a research to measure the thread depth at car tire using matlab. so far im still stuck at extraction stage. kindly give me an idea how to measure the depth on car tire. tq

[Star Strider]

What data are you starting with?

[John BG]

hang an image so we can have a look

[alef imran]

[alef imran]

[alef imran]

[alef imran]

the data from 2d images..is that possible to measure the depth? currently im using some algorithm to make the image become 3D..the picture above are the result.. but now im stuck to measure it? how about scaling the pic?

[Jon]

Maybe you set an upper threshold (around 7, say) and take the average height of all pixels greater than your threshold. This gives you a datum to work from. Then set a lower threshold (around 5, say) and create an matrix where only pixels with a height lower than 5 are included. This image should include all the treads along with the smaller isolated pock marks. Then you can do bwlabel or bwconncomp to find the largest connected blob (the treads should all be connected) and remove the smaller blobs. Finally, take the average height of that largest blob and subtract it from the average height of the datum, and that's a measure of tread depth. Looks like a fun problem!

[alef imran]

thanx.. i will try it.. its really a fun problem!

Answer 1

[Image Analyst]

How did you obtain this image? Did you use a proper profilometer to get height/depth information? Or do you have just some snapshot from your phone you took (and so are out of luck)?

Is the data in a proper image file format? If not, then please use your reading program to get it into a matrix and attach to your original image with in a .mat file. Then we can try things on it. Jon's suggestion is what I'd start with. Something like

mask = depthImage < 6;
meanDepth = mean(depthImage(mask))

If the tops are not flat then you should correct for that in your image capture setup. Otherwise you might have to estimate the tops using the mask and original image, and the roifill() function. Then you can subtract the original image from the flat top image to get better estimates of the depth in a curved-top situation.

[alef imran]

yes the data comes from snapshot from phone (front car tire view). before this im wanted to use stereo video method. but got complicated things. so this is other alternative kind of method that i wanted to try measure the depth from single camera (2D images). may i ask when to apply the suggestion after

[Image Analyst]

You can try stereo imaging if you have the Computer Vision System Toolbox. I can't help you with that since I don't have that toolbox.

Or else you can use a Kinect camera which uses time of flight. However I'm not sure what the depth resolution is. How accurately do you need to know the depth? Nearest mm? Nearest nanometer?

[alef imran]

kinect camera will give me absolute result for sure. the idea is to use other kind of method but still using matlab. just nearest mm. actually im still blur with the jon suggestion.

[Image Analyst]

Jon's method and my code will not work for a simple snapshot. It will only work for a depth image, where the pixel value represents the depth, not brightness. You're going to have to explore a stereo camera setup. Or simply use a scale or ruler for a low tech (no camera) solution.

[alef imran]

So u mean i need to calibrate the distance between camera and tire? So from there i can calculate the pixel?

[Image Analyst]

You would need that along with how separated the first camera position was from the second camera position. See this: http://www.mathworks.com/products/computer-vision/features.html#camera-calibration

[alef imran]

Ohh sure. Thanx btw. I will figure it out and share it when I succeed.