Identifying straight line & intersection with the curve from the binary image

Question

DW 2023-11-23

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2051172-identifying-straight-line-intersection-with-the-curve-from-the-binary-image

评论： DW 2023-11-29

I want to fit a blue line (curved or straight), detect a red line, and then find the slope of the light green line that is tangent to the blue line at the intersection of the two lines.

By detecting the first row in each column with the value 1, I could detect blue line.

Also, fitting blue curvature from this data would be no problem.

(Based on my previous question ( https://kr.mathworks.com/matlabcentral/answers/2050272-curvature-detection-from-outlines-of-the-binary-image?s_tid=srchtitle) and kind answer of "Image Analyst").

In the case of a straight line section, it seems to be possible to first find the row with a value of 1 for each column, average it, and then draw a straight line to the value on the leftmost side.

However, in this case, it seems that I have to manually determine the location of the pixel corresponding to the straight line section(otherwise, even curved surfaces will be recognized).

Any suggestion would be very helpful.

Thank you.

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Image Analyst 2023-11-23

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2051172-identifying-straight-line-intersection-with-the-curve-from-the-binary-image#answer_1358657

The best solution is if you can put your part into some kind of jig so that everything is in a fixed, known position. Then you can just use a fixed mask to erase everything you don't want.

Another way might be to scan the image row by row using find() and find out where that large rectangular gap is. Then just erase 3 or 4 columns just to the outside of the rectangular gap.

Another option would be to scan the image getting the top rows and then try to find the change point using findchangepts.

Another option would be to just fit everything, spike and all, then find the residuals between your fit and the actual data and say that if the residual (difference) is too much, then delete those points and train your polynomial again.