特征点检测与匹配的经典方法

特征点检测、匹配、误差剔除匹配方法python实现代码
sift

#sift
'''
import cv2
import numpy as np
import time

img = cv2.imread('11.jpg')

gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
sift = cv2.xfeatures2d.SIFT_create(500)
start = time.clock()
kp = sift.detect(gray, None)  # 找到关键点
kp_time = (time.clock() - start)
print("kp_time:", '% 4f' % (kp_time))  # ms
print(len(kp))

img = cv2.drawKeypoints(img, kp, img,(255,0,0))  # 绘制关键点
cv2.namedWindow('sift_point', 0)
cv2.imshow('sift_point', img)
cv2.imwrite('11sift.jpg',img)

cv2.waitKey(0)
cv2.destroyAllWindows()
'''

sift+bf

#sift+bf
'''
import cv2
import numpy as np
import time

# 1) 打开图像
img1 = cv2.imread("11.jpg")
img2 = cv2.imread("22.jpg")

# 2) 图像灰度化
img1_gray = cv2.cvtColor(img1, cv2.IMREAD_GRAYSCALE)
img2_gray = cv2.cvtColor(img2, cv2.IMREAD_GRAYSCALE)
start = time.clock()
# 3) SIFT特征计算
sift = cv2.xfeatures2d.SIFT_create(500)
kp1, des1 = sift.detectAndCompute(img1, None)
kp2, des2 = sift.detectAndCompute(img2, None)

bf = cv2.BFMatcher(cv2.NORM_L2)
matches1 = bf.knnMatch(des1, des2, k=2)
bf_time = (time.clock() - start)
print("bf_time:", '% 4f' % (bf_time))
# goodMatch = []
# # 匹配优化
# for m, n in matches1:
# 	# goodMatch是经过筛选的优质配对，如果2个配对中第一匹配的距离小于第二匹配的距离的1/2，基本可以说明这个第一配对是两幅图像中独特的，不重复的特征点,可以保留。
#     if m.distance < 0.6*n.distance:
#         goodMatch.append([m])
# bf_img_out = cv2.drawMatchesKnn(img1, kp1, img2, kp2, goodMatch[:], None, flags=2)
bf_img_out = cv2.drawMatchesKnn(img1, kp1, img2, kp2, matches1<