cours-m1-eea/453-Traitement_Image/Cours/hough.py

41 lines
1.5 KiB
Python
Raw Permalink Normal View History

def calcul_acc_cercles(img_s, rad_min=5, rad_max=100):
#init acc :
[c_max, r_max] = img_s.shape
r_min=1
delta_r = 1
N_r = int((r_max-r_min+delta_r)/delta_r)
delta_c = 1
c_min = 1
N_c = int((c_max-c_min+delta_c)/delta_c)
delta_rad = 1
N_rad = int((rad_max-rad_min+delta_rad)/delta_rad)
acc = np.zeros((N_rad,N_r,N_c))
print("Taille de l'accumulateur" + str(acc.shape))
x,y = np.nonzero(img_s)
for i in range(len(x)):
for r in range(r_min, r_max+1, delta_r):
for c in range(c_min, c_max+1, delta_c):
if (x[i],y[i]) != (r,c):
rad = np.sqrt((r-x[i])**2 + (c-y[i])**2)
if rad < rad_max and rad > rad_min:
i_id = int((r-r_min)/delta_r)
j_id = int((c-c_min)/delta_c)
k_id = int((rad-rad_min)/delta_rad)
acc[k_id,i_id,j_id] += 1 /rad
return acc
def cherche_N_maxima_cercles(accumulateur, exclusion_xy,
exclusion_rayon, N):
accu2 = np.copy(accumulateur)
[rad_max0, c_max0, r_max0] = accumulateur.shape
liste_max = []
for cercle in range(N):
[rayon, w_0, h_0] = np.unravel_index(np.argmax(accu2),accu2.shape)
rayon += 5
w_0 += 1
h_0 += 1
accu2[rayon - exclusion_rayon : rayon + exclusion_rayon,
w_0 -exclusion_xy : w_0 + exclusion_xy,
h_0 - exclusion_xy : h_0 + exclusion_xy] =0
liste_max.append([rayon, w_0, h_0])
return liste_max