def match_image_template(image, template_dir, thresh):
"""
A function to match the templates with the image.
Input:
image: Image from which needs to find patterns
template_dir: A directory containing (cropped) template images
thresh: Score Threshold Value.
Output: matched patterns in pandas dataframe
BBox, Score, templateName
"""
list_ = []
for i in os.listdir(template_dir):
template = cv2.imread(template_dir + i, 0)
list_.append((str(i), template))
match = matchTemplates(list_,
image,
score_threshold=thresh / 100,
method=cv2.TM_CCOEFF_NORMED,
maxOverlap=0.30)
print("Found {} Ellipses".format(len(match.index)))
print(match)
return match
def locate_particles(frame, template, match_features=features_match):
hits = matchTemplates(
template.create(),
frame,
N_object=match_features.get('N_obj'),
score_threshold=match_features.get('threshold'),
method=cv2.TM_CCOEFF_NORMED,
maxOverlap=match_features.get('max_overlap')).sort_index()
hits.index = range(
len(hits)
) #WARNING: this tris to solve the problem with matchTemplate, as it produces a dataframe with missing indices
if template.find_centres() is None:
return hits
else:
centres = template.find_centres()
hits1 = hits['BBox']
bbox = []
for i in range(len(hits)):
x = hits1[i][0] + centres.get(hits.TemplateName[i])[0]
y = hits1[i][1] + centres.get(hits.TemplateName[i])[1]
wx = hits1[i][2]
wy = hits1[i][3]
bbox.append((x, y, wx, wy))
hits.BBox = bbox
return hits
def check_coins(self):
ind = 0
max_index = len(self.coin_pos)
while ind < max_index:
coin1_pos = self.coin_pos[ind]
coin2_pos = self.coin_pos[ind + 1]
mouse_click(coin1_pos[0] + coin1_pos[2] / 2,
coin1_pos[1] + coin1_pos[3] / 2,
wait=0.1)
mouse_click(coin2_pos[0] + coin2_pos[2] / 2,
coin2_pos[1] + coin2_pos[3] / 2,
wait=0.3)
screen = cv2.imread(screen_grab())
matches = matchTemplates(self.coin_images,
screen,
N_object=2,
score_threshold=.7,
maxOverlap=.25,
searchBox=None)
coin1 = (matches["TemplateName"][0], matches["BBox"][0])
coin2 = (matches["TemplateName"][1], matches["BBox"][1])
if coin1[0] == coin2[0]:
self.coin_items.pop(coin1[0])
else:
self.coin_items[coin1[0]].append(coin1[1])
self.coin_items[coin2[0]].append(coin2[1])
ind += 2
def locate_templates(frame, list_template, match_features=features_match):
return matchTemplates(
list_template,
frame,
N_object=match_features.get('N_obj'),
score_threshold=match_features.get('threshold'),
method=cv2.TM_CCOEFF_NORMED,
maxOverlap=match_features.get('max_overlap')).sort_index()
def get_coin_fields(self):
screen = cv2.imread(screen_grab())
matches = matchTemplates(
[("card", cv2.imread("rc_items/coinflip_back.png"))],
screen,
N_object=float("inf"),
score_threshold=0.5,
#maxOverlap=0.25,
searchBox=None)
for i in range(len(matches['BBox'])):
self.coin_pos.append(matches['BBox'][i])
def find_image(image_path, root_image_path):
matches = matchTemplates(
[("img", cv2.imread(image_path))],
cv2.imread(root_image_path),
N_object=10,
score_threshold=0.9,
# maxOverlap=0.25,
searchBox=None)
if len(matches["BBox"]) == 0:
return None, None
else:
box = matches["BBox"][0]
return box[0], box[1]
def TemplateMmatching(tarImg, hostImg, thres):
target = cv2.cvtColor(tarImg, cv2.COLOR_BGR2GRAY)
listTarget = [('target', target)]
host = cv2.cvtColor(hostImg, cv2.COLOR_BGR2GRAY)
for i, angle in enumerate([90, 180]):
rotated = np.rot90(target, k=i +
1) # NB: rotate not good here, turns into float!
listTarget.append((str('target'), rotated))
hits = matchTemplates(listTarget,
host,
score_threshold=thres,
method=cv2.TM_CCOEFF_NORMED,
maxOverlap=0)
return hits
def find_point_symbol(list_template):
for template in list_template:
listTemplate = [('template', template)]
for i, angle in enumerate([90, 180]):
rotated = np.rot90(
template,
k=i + 1) # NB: np.rotate not good here, turns into float!
listTemplate.append((str(angle), rotated))
Hits = matchTemplates(listTemplate,
image_gray,
N_object=1,
score_threshold=0.6,
method=cv2.TM_CCOEFF_NORMED,
maxOverlap=0.3)
point = Hits.values.tolist()[0][1]
point_plot = (int(point[0] + (point[2] / 2)),
int(point[1] + (point[3] / 2)))
cv2.circle(image, point_plot, 2, (0, 255, 0), -1)
def find_point_symbol(image, list_template):
font = cv2.FONT_HERSHEY_SIMPLEX
image_gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
point_symbol = []
for template in list_template:
# template = cv2.resize(template, (80, 80))
# template = cv2.flip(template, 1)
# template = template[10:90, 10:90]
# cv2.imshow("thin9",template)
# cv2.waitKey(0)
listTemplate = [('template', template)]
for i, angle in enumerate([90, 180]):
rotated = np.rot90(
template,
k=i + 1) # NB: np.rotate not good here, turns into float!
listTemplate.append((str(angle), rotated))
Hits = matchTemplates(listTemplate,
image_gray,
N_object=1,
score_threshold=0.4,
method=cv2.TM_CCOEFF_NORMED,
maxOverlap=1)
point = Hits.values.tolist()[0][1]
# print(Hits)
if float((Hits.values.tolist()[0][2])) >= 0.6:
point_plot = (int(point[0] + (point[2] / 2)),
int(point[1] + (point[3] / 2)))
point_symbol.append(point_plot)
point_symbol = Sam_OOk_isas(point_symbol, thres=5)
for point in point_symbol:
cv2.putText(image, (str(point[0]) + "," + str(point[1])), point, font,
0.5, (255, 0, 0), 2)
cv2.circle(image, point, 2, (0, 255, 0), -1)
# print(point_symbol)
return point_symbol
import matplotlib.pyplot as plt
import numpy as np
fileID = r"C:\Users\MattiaV\Desktop\università\Interships\DynamicsOfGranularShapes\ParticleTracking\Images\First frames\f1.png"
image = io.imread(fileID)
temp_draft = image[300:500,500:670]
temp0 = temp_draft[41:113,48:123]
image2 = io.imread(r"C:\Users\MattiaV\Desktop\università\Interships\DynamicsOfGranularShapes\ParticleTracking\Images\First frames\f2.png")
listTemplate = []
# Initialise figure
f, axarr = plt.subplots(1,4)
for i in np.arange(0, 360, 90):
#rotated1 = t.rotate(temp0, i, preserve_range=True)
#rotated = np.rint(rotated1).astype(int) # transforms array of float into array of int
rotated = np.rot90(temp0, k=int(i/90)) # NB: np.rotate not good here, turns into float!
listTemplate.append( (str(i%120), rotated ) )
axarr[int(i/90)].imshow(rotated)
# We could also do some flipping with np.fliplr, flipud
Hits = matchTemplates(listTemplate, image2, N_object=50, score_threshold=0.55, method=cv2.TM_CCOEFF_NORMED, maxOverlap=0.22).sort_index()
Overlay = drawBoxesOnRGB(image, Hits, boxThickness=5)
plt.figure(figsize = (30,30))
plt.axis("off")
plt.imshow(Overlay)
for i, angle in enumerate([2, 5]):
rotated2 = ndimage.rotate(
Template1, 2,
reshape=False) # NB: np.rotate not good here, turns into float!
listTemplates.append((str(angle), rotated2))
axarr[i + 1].imshow(rotated2, cmap="gray")
# We could also do some flipping with np.fliplr, flipud
# In[5]:
Hits = matchTemplates(
listTemplates,
SampleImg,
method=cv2.TM_CCOEFF_NORMED,
score_threshold=0.89,
maxOverlap=0,
)
print("Found {} hits".format(len(Hits.index)))
Hits
#Is it becuase the images are not floated?
# In[6]:
#Show Image with Template Lables
get_ipython().run_line_magic('matplotlib', 'notebook')
Overlay = drawBoxesOnRGB(ColorImg, Hits, showLabel=True)
plt.imshow(Overlay)
re = cv2.imread(r'C:\Users\aminb\Desktop\FIBO\Image\symbol_module\symbol2.png',
0)
rr = cv2.imread(r'C:\Users\aminb\Desktop\FIBO\Image\symbol_module\symbol3.png',
0)
plt.show()
# 1st format the template into a list of tuple (label, templateImage)
listTemplate = [('small', template)]
# Initialise figure
# _, axarr = plt.subplots(1,3)
# axarr[0].imshow(smallCoin, cmap="gray")
for i, angle in enumerate([90, 180]):
rotated = np.rot90(template, k=i +
1) # NB: np.rotate not good here, turns into float!
listTemplate.append((str(angle), rotated))
Hits = matchTemplates(listTemplate,
image_gray,
N_object=1,
score_threshold=0.6,
method=cv2.TM_CCOEFF_NORMED,
maxOverlap=0.3)
print(Hits)
point = Hits.values.tolist()[0][1]
point_plot = (int(point[0] + (point[2] / 2)), int(point[1] + (point[3] / 2)))
cv2.circle(image, point_plot, 2, (0, 255, 0), -1)
cv2.imshow("Ki", image)
cv2.waitKey(0)
try:
image = io.imread(file, 0) ########## reading image
image = image[0:500, 165:500]
name1 = file.split("/")[-1]
name1 = name1.split(".")[0]
im = image
noise = np.empty_like(im, dtype="int8")
level = 10
cv2.randn(noise, (0), (level)) # Matrix element are 0 in average
imageNoise = cv2.add(im, noise, dtype=cv2.CV_8U)
Hits_Noise = matchTemplates(listTemplate,
imageNoise,
N_object=90,
score_threshold=0.001,
method=cv2.TM_CCOEFF_NORMED,
maxOverlap=.08)
H = Hits_Noise
df = Hits_Noise.reset_index()
w = df["BBox"].str[2]
h = df["BBox"].str[3]
df["width"] = round(w * 2.54 / 96, 1)
df["hight"] = round(h * 2.54 / 96, 1)
f1 = df.TemplateName.unique()
br1 = df['TemplateName'].value_counts().reset_index()
# resize the marker according to the scale, and keep track
# of the ratio of the resizing
new_copy = paper_gray.copy()
resized = imutils.resize(omr_marker,
width=int(omr_marker.shape[1] * scale))
r = omr_marker.shape[1] / float(resized.shape[1])
if resized.shape[0] > height or resized.shape[1] > width:
break
listTemplate = [('marker', resized)]
Hits = matchTemplates(
listTemplate,
new_copy,
N_object=4,
score_threshold=0.4,
maxOverlap=0,
method=cv2.TM_CCOEFF_NORMED) #,searchBox=(76,781,1856,353)
Overlay = paper.copy()
for index in Hits.index:
x1, y1, w, h = Hits.BBox[index]
Overlay = drawBoxesOnRGB(paper, Hits)
Sum_scores = Sum_scores + Hits.Score[index]
#print("Sum_scores : ",Sum_scores)
#cv2.imshow('template_matched_paper',cv2.resize(Overlay,(620,620)))
#cv2.waitKey(0)
import matplotlib.pyplot as plt
import numpy as np
image = "Testcopy.png"
plt.figure(figsize=(10, 10))
plt.axis("off")
plt.imshow(image, cmap="gray")
temp0 = image[784:784 + 400, 946:946 + 414] # with well 49
plt.axis("off")
plt.imshow(temp0, cmap="gray")
listTemplate = [('temp0', temp0)]
Hits = matchTemplates(listTemplate,
image,
N_object=4,
score_threshold=0.3,
method=cv2.TM_CCOEFF_NORMED,
maxOverlap=0.3)
print(Hits)
# Generate gaussian distributed noise, the noise intensity is set by the level variable
noise = np.empty_like(image, dtype="int8")
level = 50
cv2.randn(noise, (0), (level)) # Matrix element are 0 in average
plt.figure(figsize=(10, 10))
plt.axis("off")
plt.imshow(noise, cmap="gray")
imageNoise = cv2.add(image, noise, dtype=cv2.CV_8U)
im = image
#Hits = matchTemplates(listTemplate, im,N_object=34, score_threshold=0.5, method=cv2.TM_CCOEFF_NORMED, maxOverlap=0.25)
#print(Hits)
# Generate gaussian distributed noise, the noise intensity is set by the level variable
noise = np.empty_like(im, dtype="int8")
level = 10
cv2.randn(noise, (0), (level)) # Matrix element are 0 in average
imageNoise = cv2.add(im, noise, dtype=cv2.CV_8U)
Hits_Noise = matchTemplates(
listTemplate,
imageNoise,
score_threshold=0.75,
method=cv2.TM_CCOEFF_NORMED,
maxOverlap=0.2,
searchBox=None) ##########without using N_object
#Hits_Noise = matchTemplates(listTemplate, imageNoise,N_object=20,score_threshold=0.001, method=cv2.TM_CCOEFF_NORMED, maxOverlap=.08,searchBox=None)
H = Hits_Noise
df = Hits_Noise.reset_index()
w = df["BBox"].str[2]
h = df["BBox"].str[3]
df["width"] = round(w * 2.54 / 96, 1)
df["hight"] = round(h * 2.54 / 96, 1)
#score=max(df.Score)
image_nored = removeRedRegionFromImage(BGR_cropimg, mask_closedred)
cv2.imshow('No red white region', image_nored)
cv2.waitKey(1)
#In[4]: APPLY TEMPLATE MATCHING WITHOUT RED REGION
listTemplate_inwhite = [('A', BGR_featureimg_kngp),
('B', BGR_featureimg_kngp_2),
('G', BGR_featureimg_kngp_3),
('D', BGR_featureimg_kngp_4)
] #,('E', BGR_featureimg_kngp_5)]
Hits = matchTemplates(listTemplate_inwhite,
image_nored,
score_threshold=0.55,
method=cv2.TM_CCOEFF_NORMED,
maxOverlap=0)
Overlay_inwhite = drawBoxesOnRGB(image_nored, Hits, showLabel=True)
total_inwhite = len(Hits)
listTemplate_blueinwhite = [('*A', BGR_featureimg_kngn),
('*B', BGR_featureimg_kngn_2),
('*C', BGR_featureimg_kngn_3)]
Hits_blueinwhite = matchTemplates(listTemplate_blueinwhite,
image_nored,
score_threshold=0.75,
method=cv2.TM_CCOEFF_NORMED,
maxOverlap=0)
Overlay_blueinwhite = drawBoxesOnRGB(image_nored,
from skimage.data import coins
import matplotlib.pyplot as plt
image = coins()
plt.imshow(image, cmap="gray")
smallCoin = coins()[37:37 + 38, 80:80 + 41]
plt.imshow(smallCoin, cmap="gray")
# 1st format the template into a list of tuple (label, templateImage)
listTemplate = [('small', smallCoin)]
# Then call the function matchTemplates (here a single template)
Hits = matchTemplates(listTemplate,
image,
score_threshold=0.5,
method=cv2.TM_CCOEFF_NORMED,
maxOverlap=0)
print("Found {} hits".format(len(Hits.index)))
Overlay = drawBoxesOnRGB(image, Hits, showLabel=True)
plt.imshow(Overlay)
Hits = matchTemplates(listTemplate,
image,
score_threshold=0.4,
method=cv2.TM_CCOEFF_NORMED,
maxOverlap=0)
Overlay = drawBoxesOnRGB(image, Hits, showLabel=True)
plt.imshow(Overlay)
