How to apply Matlab CNN code to analyze the image?

Hi.

Using Matlab, we want to analyze the proportions of the elements inside the circle except for the outer part of the circle in the picture. The photo is part of the asphalt cross section taken by CT. The method I tried was through histogram analysis,

The black part is separated by air

If the remaining white and gray remain, the color from white to gray will be classified as aggregate (stone), and the remaining gray will be classified as asphalt (adhesive-like material).

At first, I didn't know how to do matlab at all.

I ran into a limitation in writing code and wrote to fix this.

Also, because I have thousands of related pictures

You may be able to solve the problems mentioned above through learning (CNN).

I would be very grateful if you could write some code or give advice on this.

For example, using a lot of CT data

The black part is classified as air,

White parts and intermediate colors that go from white to gray are classified as aggregates,

The gray part is classified as asphalt (adhesive-like material)

I want to find out the ratio.

My preferred method is to use CNN. In addition, I would like to see classified air, aggregate, and asphalt separately on the matlab. Parts outside the circle are always excluded from analysis.

Answer 1

[Yogesh Khurana]

Hi,

I am assuming that you want to create a new model for classifying the regions in attached image or similar to that. I would suggest you some links that can be useful for creating a CNN model in MATLAB and using that for object (region) classification (I think you can't use pretrained models here). Please refer to the following links for more information:

https://www.mathworks.com/help/deeplearning/examples/create-simple-deep-learning-network-for-classification.html

https://www.mathworks.com/help/deeplearning/gs/create-simple-deep-learning-classification-network.html

https://www.mathworks.com/help/deeplearning/deep-learning-with-images.html

Hope it will help!