I would like to use find_contours and regionprops of scikit-image (also opencv2 module is good) python module to extract feature from a binary image. It's not clear how can i watch contours founded by find_contours. For example: if i have in an image 1 circle, a triangle and a casuale line, how can i select firt two figures to calculate area and perimeter, and select third to calculate it's orientation?
Then, which are rotation and scale invariant properties?
All the properties of a single object, i thought to put in an array (with an id identifier and the image name which owns) and for making comparison among them use a Support Vector Machine, could be correct?
a sample image: http://www.sistemamuseo.it/data/immagini_servizieducativi/20120404024953_img_d4459.jpg (usually there are not so many symbol in a single image)