用于图像识别的模式识别

Question

I need to write a program that recognize some patterns in different photos. I wrote a program that take a photo as input and creates another image with the edges of previous photo. Now I'm stuck with pattern detection. I tried to take 2d arrays of pixel and mark each possible pattern by giving each pixel a value from 0 to n(maximum number of pixel in a sequence). Then I take the objects that the program already knows and see which one contains the more patterns that were found.

The problem is that beside efficiency, the program won't work if the image is upside down(If I train it with a photo and then flip the photo, the program won't recognize it).

Can you tell me some methods to fulfill my task, or some good tutorials or courses that explain the process a lot deeper than just:"search for patterns"?

Answer 1

Your problem description is very general. To get better answers provide some input data characteristics as well as describe what kind of patterns you're looking for.

What could be useful in general problem of pattern recognition, is using neural networks. For example you could check the first chap. of this book http://neuralnetworksanddeeplearning.com/chap1.html

There's simple example of pattern recognition for handwritten digits. In your case for solving rotation problem, you'd probably have to rotate training example as well.