Coins detection¶

Setup¶

from PIL import Image
import numpy as np
from matplotlib import pyplot as plt
from matplotlib import cm
from scipy import ndimage

%matplotlib nbagg

Open the file¶

!ls

balls.jpg  Image_Processing_Tutorial_1.ipynb  Image_Processing_Tutorial_4.ipynb
bugs.jpg   Image_Processing_Tutorial_2.ipynb  Orientation.ipynb
coins.jpg  Image_Processing_Tutorial_3.ipynb

im = Image.open('coins.jpg')
channels = im.split()
r = np.array(channels[0])

fig = plt.figure("Red Channel")
plt.imshow(r, origin = "upper", cmap = cm.gray)
plt.colorbar()
plt.show()

<IPython.core.display.Javascript object>

Plot the histogram¶

fig = plt.figure("Histogram")
plt.hist(r.flatten(), bins=np.arange(256), histtype = "stepfilled")
plt.show()

<IPython.core.display.Javascript object>

Thresholding¶

seuil = 80
ims = r > seuil
binaire = np.where(ims,1,0)

fig = plt.figure("Thresholding")
plt.imshow(binaire, origin = "upper", cmap = cm.gray)
plt.colorbar()
plt.show()

<IPython.core.display.Javascript object>

Erosion + Labeling¶

erod = ndimage.morphology.binary_erosion(binaire, structure=np.ones((10,10)))
lab, number = ndimage.measurements.label(erod)

fig = plt.figure("Erosion effect")
ax = fig.add_subplot(1, 3, 1)
ax.axis("off")
plt.title("Erosion ON")
plt.imshow(erod, origin = "upper", cmap = cm.binary)
ax = fig.add_subplot(1, 3, 2)
ax.axis("off")
plt.title("Erosion OFF")
plt.imshow(binaire, origin = "upper", cmap = cm.binary)
ax = fig.add_subplot(1, 3, 3)
ax.axis("off")
plt.title("Labeling")
plt.imshow(np.where(lab,lab,np.nan), origin = "upper", cmap = cm.jet)
#plt.colorbar()
plt.show()

print(number)

<IPython.core.display.Javascript object>

fig = plt.figure("Labeling")
plt.title("Nombre de grains")
plt.imshow(np.where(lab,lab,np.nan), origin = "upper", cmap = cm.inferno)
plt.colorbar()
plt.show()

<IPython.core.display.Javascript object>

Sobel filter¶

sobx = ndimage.sobel(erod, axis=0, mode='constant')
soby = ndimage.sobel(erod, axis=1, mode='constant')
sob = np.hypot(sobx, soby)
bins = 1000

fig = plt.figure("Sobel–Feldman operator")
plt.imshow(np.where(sob > 0 , bins, r), origin = "upper",cmap = cm.gray)
plt.show()

<IPython.core.display.Javascript object>

Localization of centers¶

fig = plt.figure("Centers")
plt.imshow(np.where(erod == 1, r, np.nan), origin = "upper")
ax = fig.add_subplot(1,1,1)
#ax.axis("off")
for i in range(1,np.max(number)+1):
    x,y= np.nonzero(np.where(lab==i,1,0))
    xmean=x.mean()
    ymean=y.mean()
    plt.plot(ymean,xmean,"k+-", markersize=60)

plt.show()

<IPython.core.display.Javascript object>

fig = plt.figure()
plt.imshow(r, origin = "upper")

centers = []
for l in np.arange(1, number+1):
        y, x = np.where(lab == l)
        centers.append([x.mean(), y.mean()])
centers = np.array(centers).T
plt.plot(centers[0], centers[1], "or")
plt.show()

<IPython.core.display.Javascript object>

np.unique(lab)

array([0, 1, 2, 3, 4, 5, 6, 7])