How to implement a funtion to 'Count an amount of one type of lego piece in the Image'?

Sorry if it's a lot to ask but is there an easy way to implement a function that takes in the image of cluster of lego pieces and returns the count of a particular type of Lego piece? i.e. given the lego piece type shown in the first image--------->

how to implement a function that return the amount of the same type of lego pieces existed in the image below.

Thank you very much in advance!!

Baz

Answer 1

[Image Analyst]

First you have to do color segmentation to get separate, individual blocks. See my File Exchange for demos: http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862

Then try using imfindcircles() to count the circles in each colored brick. Here is a demo: http://www.mathworks.com/matlabcentral/fileexchange/34365-circle-finder

[Chanodom Utsahajit]

Thank You very much!, This helped me a lot :D

[Image Analyst]

You're welcome. Can you mark the answer as accepted then? Thanks in advance. And you can come back with your program and images once you're more into the project. You could make your job a lot easier by having your camera directly overhead and using flood, "integrating sphere" type of lighting to eliminate shadows.

[Nafis Abdullah Bin Azad Akter]

@ImageAnalyst

You cant count circles for pieces that are turned over, how would you count the circles of pieces that dont show the circles, instead is there no way to find pieces by the length of their longest edge, and also how could you detect objects by edge when an image has been binarized. And those images we cant change them they were given to us in that way and we have to be able to detect the pieces of specific size and color. For example a 2x2 piece that is blue and we need to count how many of those 2x2 pieces are in the test images given to us. Our objective is to research online and find a way to do that sole thing.

Kind regards,

Nafis

[Image Analyst]

I doubt you'd be able to do research online and find a solution like that, unless you're very very lucky. I think you're going to be stuck writing code.

Even so, there are cases when you just can't get it right. For example, in a jumbled pile of Lego's, if you see just the end of a block, there is no way of knowing if it's a 4-bump Lego or an 8-bump Lego. No way. They both look identical.

Good luck though.

[Nafis Abdullah Bin Azad Akter]

I know there is no way but is there any way to measure the length of the side of the pieces thats longest along the perimeter. And if there is no way to do that is there a way to measure at least the bounding boxes longest side. Or get the bound area to a threshold measuring the ratio of the other pieces and create a ratio of the smallest section and the biggest area, and discard only the small sectors, Areafilter has an option to get the biggest chunks in an image but it has a N number input how would someone go on to make it a variable amount depending on image. Or measure from a centroid to the corner of the object at least. Would ConvexHull property from regionprops be of any use in this task. Or instead would cuboid search with regionprops3 be of any use for the task. The following images illustrate the bounding boxes i created around the objects but cannot measure them just yet. Thanks for your help in advance.

Kind regards,

Nafis

[Image Analyst]

Well of course you can ask regionprops() for the BoundingBox, but I think you may want the Feret Diameters.

See this link: Image Processing Blog on Feret Diameters

[Shruti Jha]

Adding on to this thread, I have a similar problem. I am able to isolate blue coloured legos, however when I perform edge detection and filling, I get something like this (the last image).

Now, I'm not sure how to fill all the others and separate out the rectangular ones. Can somebody please help?

Much appreciated!