SurfaceDefects Python Package

A simple but powerful Python package capable of

1. detecting pores in concrete, steel and similar surafaces, or the fabrication errors for that matter.
2. determining the area of each pore.
3. calculating equivalent diameter of each pore. Equivalent diameter is the diameter of a circle whose area equals that of a pore. More on it later...
4. calculating surface porosity i.e., the ratio of area covered by pores to the total area.
5. plotting circles over pores once the equivalent diameters are determined.
6. assigning length scale to image.

It works on the spirit of Computer Vision and segmentation. The Python package SurfaceDefects classifies surfaces based on the difference in their pixel values. It can also generally be used for detecting contours in 2D images of any type provided they are suitable for the intended purpose.

Better images lead to better results!

The concept will develop as we go down the documentation.

1. Install SurfaceDefects

~ pip install SurfaceDefects

2. Import into the project alongwith matplotlib.pyplot

import SurfaceDefects as sdf
import matplotlib.pyplot as plt

3. Load image using LoadImage() function

img = sdf.LoadImage(path, gray=Fasle) 

path: is the complete path to your image

gray: whether to load the image as a grayscale image, default is False

Tip: Load img as it is and another img_gray as a graysacle image. You will need both of them later.

4. Make a black replica of img (not compulsory)

black = sdf.PseudoImage(img) 

5. Detect pores/defects using ClassifySurface() function

thresh = sdf.ClassifySurface(img_gray, threshold, invert_color=True)

plt.imshow(thresh, cmap='gray')

thresh: the resultant image is better known as thresholded image hence thresh

img_gray: the original grayscale image

threshold: verily the most important parameter, threshold is what that controls the detection of pores/defects. Ranges between 0 (black) and 255 (white).

invert_color: swaps colors i.e., black to white. Default is True

---

plt's cmap='gray' for a grayscale image.

6. Find boundary of pores/defects using FindContours() function

Contours are lines joining points of same instensity.

Those points in this case are located along the boundaries of pores.

contours, _ = sdf.FindContours(classified_img)

contours: an array

_: the hierarchy and can be ignored in this particular case.

classified_img = thresh, the thresholded image.

7. SignificatContours()

Some of the contours found by FindContours() might be insignificant in that their areas might be zero. SignificatContours() will automatically discard contours whose areas are zero.

sig_contours, _ = sdf.SignificantContours(classified_img)

sig_contours: an array

_: the hierarchy and can be ignored in this particular case.

classified_img = thresh, the thresholded image.

8. Define a scale using DefineScale() function

scale_factor = sdf.Scale(height, width) 

scale_factor: a ratio of scale, or better known as a Referesentative Factor

height, width: height and width of real object/surface in units of length

9. Calculate area of pores using ContourArea()

pore_area = sdf.ContourArea(classified_img, sum=False, scale_factor=1)

If sum=True, total pore area will be calculated; otherwise a list.

scale_factor: a ratio of scale, or better known as a Referesentative Factor. Default value is $1$ which means no conversion from units of pixel to units of length.

10. EquivalentDiameter() calculates diameter of a circle whose area equals that of a pore

equ_dia =  sdf.EquivalentDiameter(contours, scale_factor=1)

11. DrawContours() draws boundaries of each pore

contour_img = sdf.DrawContours(img,contours, index=-1, color = (0,255,0), thickness=1)

#display
plt.imshow(contour_img)

img: original or black img

contours: array from DrawContours() or SignificantContours()

index: which contour boundary to plot? $-1$ means all. Similarly $2, 3, 4$ for number $3$^rd, $4$^th, $5$^th respectively, and so on.

color: color of boundary line, default is green.

thickness: thickness of boundary line, default is 1.

12. DrawCircles() draws circle aroud the boundary of equivalent diameter

circle_img = sdf.DrawCircles(img, contours, color=(0,255,0), thickness=1)

plt.imshow(circle_img)

Refer to point number $11$ for parameters/arguments.

13. Denoise() smoothens/denoises the original image

#applies GaussianBlur filter of opencv-python

denoised_img = sdf.Denoise(img, kernel=(3,3), channels=0)

kernel: default ($3,3$)

channels: $3$ for an RGB/BGR image and $0$ for a grayscale image

14. SurfacePorosity() calculates the ratio in percentage of area covered by pores to the total area

porosity = sdf.SurfacePorosity(img, contour_area)

contour_area: a float or an integer

---

Acknowledgement | Dependencies

1. opencv-python
2. numpy