17/07/2024, 23:11 SIFT feature matching
SIFT Feature Matching
In [2]: # Import modules
import cv2 as cv
import numpy as np
import matplotlib.pyplot as plt
# Load images
image = cv.imread('store/store_shelf.png', cv.IMREAD_REDUCED_COLOR_2)
assert image is not None
template = cv.imread('store/oat_crunch.jpg', cv.IMREAD_REDUCED_COLOR_2)
assert template is not None
In [3]: # Create Sift routine
sift = cv.SIFT_create()
kp1, des1 = sift.detectAndCompute(image, None)
kp2, des2 = sift.detectAndCompute(template, None)
In [4]: print(f"Number of keypoints in first image: {len(kp1)}")
print(f"Number of keypoints in second image: {len(kp2)}")
Number of keypoints in first image: 2409
Number of keypoints in second image: 5864
In [19]: # Flaan matcher
FLAAN_INDEX_KDTREE = 1
index_params = dict(algorithm=FLAAN_INDEX_KDTREE, trees=5)
search_params = dict(checks=50)
flann = cv.FlannBasedMatcher(index_params, search_params)
matches = flann.knnMatch(des1,des2,k=2)
# Need to draw only good matches, so create a mask
matchesMask = [[0,0] for i in range(len(matches))]
# Ratio test as per Lowe's paper
for i,(m,n) in enumerate(matches):
if m.distance < 0.7*n.distance:
matchesMask[i]=[1,0]
dra_params = dict(matchColor = (0,255,0),
singlePointColor = (255,0,0),
matchesMask = matchesMask,
flags = cv.DrawMatchesFlags_DEFAULT)
# Draw matches
img3 = cv.drawMatchesKnn(image, kp1, template, kp2, matches, None, **dra_
plt.figure(figsize=(10, 8))
plt.imshow(cv.cvtColor(img3, cv.COLOR_BGR2RGB))
plt.axis('off')
plt.show()
localhost:8888/nbconvert/html/SIFT feature matching.ipynb?download=false 1/4
17/07/2024, 23:11 SIFT feature matching
Trying this out for other images
In [21]: # Load images
image = cv.imread('books.jpg', cv.IMREAD_REDUCED_COLOR_2)
assert image is not None
template = cv.imread('atomic_habits.jpg', cv.IMREAD_REDUCED_COLOR_2)
assert template is not None
In [40]: # Plot images side by side
plt.figure(figsize=(15, 7))
# Plot left image
plt.subplot(1, 2, 1)
plt.imshow(cv.cvtColor(image, cv.COLOR_BGR2RGB))
plt.title('Image')
plt.axis('off') # Hide the axes
# Plot right image
plt.subplot(1, 2, 2)
plt.imshow(cv.cvtColor(template, cv.COLOR_BGR2RGB))
plt.title('Template')
plt.axis('off') # Hide the axes
plt.show()
localhost:8888/nbconvert/html/SIFT feature matching.ipynb?download=false 2/4
17/07/2024, 23:11 SIFT feature matching
In [35]: # Create Sift routine
sift = cv.SIFT_create()
kp1, des1 = sift.detectAndCompute(image, None)
kp2, des2 = sift.detectAndCompute(template, None)
In [23]: print(f"Number of keypoints in first image: {len(kp1)}")
print(f"Number of keypoints in second image: {len(kp2)}")
Number of keypoints in first image: 1560
Number of keypoints in second image: 33938
In [24]: # Flaan matcher
FLAAN_INDEX_KDTREE = 1
index_params = dict(algorithm=FLAAN_INDEX_KDTREE, trees=5)
search_params = dict(checks=50)
flann = cv.FlannBasedMatcher(index_params, search_params)
matches = flann.knnMatch(des1,des2,k=2)
# Need to draw only good matches, so create a mask
matchesMask = [[0,0] for i in range(len(matches))]
# Ratio test as per Lowe's paper
for i,(m,n) in enumerate(matches):
if m.distance < 0.7*n.distance:
matchesMask[i]=[1,0]
dra_params = dict(matchColor = (0,255,0),
singlePointColor = (255,0,0),
matchesMask = matchesMask,
flags = cv.DrawMatchesFlags_DEFAULT)
# Draw matches
img3 = cv.drawMatchesKnn(image, kp1, template, kp2, matches, None, **dra_
plt.figure(figsize=(10, 8))
plt.imshow(cv.cvtColor(img3, cv.COLOR_BGR2RGB))
plt.axis('off')
plt.show()
localhost:8888/nbconvert/html/SIFT feature matching.ipynb?download=false 3/4
17/07/2024, 23:11 SIFT feature matching
In [ ]:
localhost:8888/nbconvert/html/SIFT feature matching.ipynb?download=false 4/4
