Formally, a computer program should be able to scan such an image, perform image processing or any specific treatment on it and produce the followings.
1) The info about multiple geometrical shapes stacked together in front/side and top views in the image.
2) The correlation between the two views as some shapes (or a part of a shape) are hidden in one view but their projections are seen in the other view.
3) The relationship among shapes such as the orientation of a shape w.r.t. to each other.
4) The info about dimensions normally written in text besides arrows.
5) The info about arrows, single-headed, double-headed, straight, slanted, etc.
Thanks Rinat for a speedy response and for so many resources. My apologies for not writing about what inputs I have to do such engineering drawing image analysis. Well, I have a small collection of some sample shapes, may be a few different variations of it, but certainly not a huge dataset to apply ML, ANN, DL, CNN. Actually I want to initially stay away from any training kind of thing or using any learning technique. I want to focus purely on image processing techniques to solve my problem and if need arises I might look into AI, etc. in a later stage. Yes, for 5th question arrows can be considered as yet another shapes. So in addition to all the resources you have given above if you can provide some pointers for image processing techniques for this problem domain I would highly appreciate it.
Hi Rinat, really thanks for providing so many pointers and sharing so many articles. I will look into each of them one by one and see what works for me. I will keep you updated here and if I get stuck with any doubt I will approach you again. May I ask, are you into image processing and CV at UoW? Thx
Thanks Fabio for the answer and particularly referring me towards your own work. That's the reason I like RG very much; I always get current state-of-the-art replies from the research community. I will mention your work in my implementation. Coding is not a problem as I usually dabble hands in Python, R on one side and Matlab on the other side. What I am looking for are the replies from experts like you who can advise what techniques can/should work and what won't work so that I can narrow down my focus and avoid wasting too much time on exploring/reading the literature. On the other hand I would like to draw your attention towards the attached engineering drawing image. Do you think after applying edge detection to this image I can use your work to recognize the primitive shapes involved and their orientation w.r.t. each other? By orientation I mean position of each shape considering the image in a 2-D plane. I have this image and a excel data file which contains info about shapes such as name of the shape and its dimensions. Thanks a lot for your replies.
Thanks Fabio for explaining the process to apply your methods. I would ask do you have any documentation or examples that you developed yourself in your work which I can read or use towards my understanding? Some usage examples would be really helpful. Also, is it possible to connect with you via email for one-to-one communication?
- Badges
- Science topic
- Similar topics
- Computer Science
- Images