The main program is main.py. This is the program launcher that adjusts the function when the sample image is loaded.
The next programs are Axis_looker.py with axillary rotate_bound.py. The secondary program for further analysis is data_read.py.
Axis_looker takes as an argument path to an image and returns an array of four arrays each containing image, x coordinate of central point of the plot, y coordinate of centre of plot. The programme finds all lines (using Hough lines transformation) on the image and selects the longest perpendicular as an axis then propagates the image and found axes to rotate_bound which output of rotated images is then returned.
Rotate_bound find crossing points of axes and angle of rotation from vertical -horizontal position. The crossing point of axes is a point around which the image is then rotated using an appropriate rotation matrix. Rotate_bound returns an array of four arrays containing rotated images and central points (with already rotated coordinates).
data_read takes 3 arguments image, x and y coordinates of the crossing point of axes and returns image and coordinates of plot function points. Program converts copy of the image to grey scale then masks all points in specific grey scale (get rid of bright colour points). Then reads all remaining points position apart of points around axes. Later it dots the original image at all found points blue and then returns an array consisting of an image and a list of plot points.
How to use the program?
- Go to the main folder,
- Run main.py in python (./main.py),
- Load a picture from the examples folder,
- Enjoy the program and its results.
We are using: PyQt5 https://pypi.org/project/PyQt5/
Sys https://docs.python.org/3/library/sys.html
Webbrowser https://docs.python.org/3/library/webbrowser.html
Os https://docs.python.org/3/library/os.html
Cv2 https://pypi.org/project/opencv-python/
The main menu of the program from which the image is loaded and then the result is obtained in the form of the output image.
We are using:
NumPy https://numpy.org/
OpenCv https://docs.opencv.org/4.x/d6/d00/tutorial_py_root.html
math https://docs.python.org/3/library/math.html.
Image is converted to gray scale and then using OpenCv bitwise_and https://docs.opencv.org/4.x/d0/d86/tutorial_py_image_arithmetics.html, program show only points from certain gray scale.
Then remaining points coordinates are then saved using OpenCv findNonZero, but points close to axes are excluded. Then the rotated image is ploted with blue dotes in position of saved coordinates.
We are using:
NumPy https://numpy.org/
OpenCv https://docs.opencv.org/4.x/d6/d00/tutorial_py_root.html
math https://docs.python.org/3/library/math.html.
Using edge points of lines program determines if axes need rotation (check if one is vertical) and finds crossing point of axes and angle of rotation.
New width, height and center point are calculated. Finally using OpenCV getRotationMatrix2D and warpAffine https://docs.opencv.org/3.4/d4/d61/tutorial_warp_affine.html rotation is made. Note that there are 4 possibilities of correct rotation each of them is tilted by multiplicity of 90 degres from another. Function returns array of four array each containing image and point of axes crossing after rotation.
We are using:
NumPy https://numpy.org/
OpenCv https://docs.opencv.org/4.x/d6/d00/tutorial_py_root.html
math https://docs.python.org/3/library/math.html.
The program load an iamge, then detects edges and convert it to a gray scale using cv2.Cannyhttps://docs.opencv.org/4.x/da/d22/tutorial_py_canny.html. Then using OpenCv HoughLinesP https://docs.opencv.org/3.4/d9/db0/tutorial_hough_lines.html program detects and stores lines. (Note that program uses probabilistic Hough transform) For machine resources resons algoritm only work if HoughLinesP didn't detect too many lines. If this happend You may manually change treshold for detection in HoughLinesP or increase accepted number of lines.
In teh next step program checks the lines, looking for pair of perpendicular lines and simulanously chacking if the line isn't image boundary. The perpendicularity is checked using scalar product. Axes are chosen as longest (sum of lenghts of lines) perpendicular line pair. Then the lines with original image are forwarded to rotate bound.