Hi Aiora,

Perhaps what you are looking for is called template matching. This functionality is available in OpenCV (C++ but has Python bindings) and the scikit-image python package. Otherwise you may consider training a neural network for the purpose.

OpenCV example: https://docs.opencv.org/master/d4/dc6/tutorial_py_template_matching.html

If the shape of the object depends on few parameters, you could use generalized Hough transformation. In order to increase speed, you could apply some edge detection before. Generalized Hough transformation can be used for circles, simple polygons, and of course for detecting lines.

Another method comes from OCR. Objects in the plane can be likened to letters. Compute the histogram of edge gradients, it will give you enough information to identify shapes. Or the histogram of distances between every two points on the border.

Or apply suited convolution operations to locate prominent points, and investigate the distances between them in the neighborhood. For example, if you want to find out whether points form a square grid, check whether neighbors of detected points are in 1 and in 1.41 unit distances.

Regards,

Joachim

Dear Aiora Zabala .

In my point of view, the python language has a lot of benefits in Image processing. For example, you can find great libraries that make you free of coding. However, if I were you, I would invest in my personal code.

Some libraries:

Scikit-image

OpenCV

Mahotas

SimplelTK

SciPy

Pillow