模板匹配和卷积原理很像,模板在原图像上从原点开始滑动,计算模板与(图像被模板覆盖的地方)的差别程度,这个差别程度的计算方法在opencv里有6种,然后将每次计算的结果放入一个矩阵里,作为结果输出。假如原图形是AxB大小,而模板是axb大小,则输出结果的矩阵是(A-a+1)x(B-b+1)
TM_SQDIFF:计算平方不同,计算出来的值越小,越相关
TM_CCORR:计算相关性,计算出来的值越大,越相关
TM_CCOEFF:计算相关系数,计算出来的值越大,越相关
TM_SQDIFF_NORMED:计算归一化平方不同,计算出来的值越接近0,越相关
TM_CCORR_NORMED:计算归一化相关性,计算出来的值越接近1,越相关
TM_CCOEFF_NORMED:计算归一化相关系数,计算出来的值越接近1,越相关
公式:https://docs.opencv.org/3.3.1/df/dfb/group__imgproc__object.html#ga3a7850640f1fe1f58fe91a2d7583695d
# 模板匹配 img = cv2.imread(‘lena.jpg‘, 0) template = cv2.imread(‘face.jpg‘, 0) h, w = template.shape[:2] img.shape template.shape
效果:
(263, 263)
(110, 85)
res = cv2.matchTemplate(img, template, cv2.TM_SQDIFF) #也可以写数字 res.shape
效果:
(154, 179)
min_val, max_val, min_loc, max_loc = cv2.minMaxLoc(res) #最大值、最小值、最大值坐标、最小值坐标 min_val max_val min_loc max_loc
效果:
39168.0 74403584.0 (107, 89) (159, 62)
for meth in methods: img2 = img.copy() # 匹配方法的真值 method = eval(meth) #print (method) res = cv2.matchTemplate(img, template, method) min_val, max_val, min_loc, max_loc = cv2.minMaxLoc(res) # 如果是平方差匹配TM_SQDIFF或归一化平方差匹配TM_SQDIFF_NORMED,取最小值 if method in [cv2.TM_SQDIFF, cv2.TM_SQDIFF_NORMED]: top_left = min_loc else: top_left = max_loc bottom_right = (top_left[0] + w, top_left[1] + h) # 画矩形 cv2.rectangle(img2, top_left, bottom_right, 255, 2) plt.subplot(121), plt.imshow(res, cmap=‘gray‘) plt.xticks([]), plt.yticks([]) # 隐藏坐标轴 plt.subplot(122), plt.imshow(img2, cmap=‘gray‘) plt.xticks([]), plt.yticks([]) plt.suptitle(meth) plt.show()
效果:
img_rgb = cv2.imread(‘mario.jpg‘) img_gray = cv2.cvtColor(img_rgb, cv2.COLOR_BGR2GRAY) template = cv2.imread(‘mario_coin.jpg‘, 0) h, w = template.shape[:2] res = cv2.matchTemplate(img_gray, template, cv2.TM_CCOEFF_NORMED) threshold = 0.8 # 取匹配程度大于%80的坐标 loc = np.where(res >= threshold) for pt in zip(*loc[::-1]): # *号表示可选参数 bottom_right = (pt[0] + w, pt[1] + h) cv2.rectangle(img_rgb, pt, bottom_right, (0, 0, 255), 1) cv2.imshow(‘img_rgb‘, img_rgb) cv2.waitKey(0)
效果:
原文:https://www.cnblogs.com/SCCQ/p/12293104.html