How to recognize the junction of walls and ceiling in a room?

Question

Pattern recognition

Hello, I’m interested in what methods can be used to recognize the line of junction of walls and ceiling in a room without resorting to neural networks? We are talking about simple walls and ceilings, without columns and other things. I'm interested in the method itself.
Using this image as an example, how to recognize seam lines (red line)

Does anyone have any ideas?

Answer 1

https://stackoverflow.com/questions/2596722/is-the...

Answer 2

You need to know the angle of view of the camera and the angle of distortion of the perspective of the camera.
We keep the constant that the angles between surfaces are always 90 degrees!
Further - we find any straight line (the junction of any two surfaces) and check by the system of equations all adjacent to it.

Answer 3

You can take a photo from the above frame and convert it to a grayscale image. After that, go through the image line by line, replace the shades of gray with the value of the gradient between the color of the current pixel and the next one (select the pixel comparison step experimentally). After that, remove from the image (replace with white color) all pixels for which the gradient is less than a certain threshold (you also need to select).
As a result, contour lines and some amount of noise will remain in the picture. Next, you need to go through the image and look for the resulting lines, each one should be entered into an array. After receiving an array of vectors, you can already put all sorts of tests with equations. This option has a nuance, all contrasting things will get into the result, for example, the contours of doors or furniture, if any.