Disclaimer: I am aware that this is a vague question, I am merely looking for suggestions on where to start. If you find that this question is better asked elsewhere, suggestions are more than welcome.
Problem: From a 2D image displaying an object (displayed as either lines or filled areas), I need to recognise and pinpoint the location of key elements such as sharp corners, rounded corners (fillets), holes, etc. in order to create a technical CAD drawing of the object.
Current knowledge: I am capable of automatically creating drawings (measurements, lines) in CAD software such as CATIA. It is the classification and location of these key features I am after.
Question(s): Simply put: where do I start? Is computer vision my answer?
I am aware of OpenCV’s contour and hough circle methods, and have tried these with a certain amount of success. This enable me to find rounded corners and sharp, 90° angles. Is OpenCV the best way to go, could you suggest different libraries, or should I ignore computer vision altogether and focus on training a convolutional neural network to classify parts of the image and go from there?
Again, apologies for the vagueness of this question from a first-timer.