opencvpython模板匹配

Template Matching

理论

模板匹配是一种在较大的图像中搜索和查找模板图像位置的方法。OpenCV带有一个函数cv2.matchTemplate()用于此目的.它只是简单地将模板图像放在输入图像上(就像在2D卷积中那样)&＃xff0c;并在模板图像下对输入图像的模板和补丁进行比较,在OpenCV中实现了几种比较方法,它返回一个灰度图像&＃xff0c;每个像素表示该像素区域与模板的匹配程度.

如果输入图像的大小&＃xff08;W x H&＃xff09;且模板图像的大小&＃xff08;w x h&＃xff09;&＃xff0c;则输出图像的大小为&＃xff08;W-w &＃43; 1&＃xff0c;H-h &＃43; 1&＃xff09;.获得结果后&＃xff0c;可以使用cv.minMaxLoc&＃xff08;&＃xff09;函数查找最大/最小值的位置。将其作为矩形的左上角&＃xff0c;并将&＃xff08;w&＃xff0c;h&＃xff09;作为矩形的宽度和高度.

OpenCV中的模板匹配

import cv2
import numpy as np
from matplotlib import pyplot as pltimg &＃61; cv2.imread(&＃39;img.jpg&＃39;,0)
img2 &＃61; img.copy()
template &＃61; cv2.imread(&＃39;img_roi.png&＃39;,0)
w, h &＃61; template.shape[::-1]# All the 6 methods for comparison in a list
methods &＃61; [&＃39;cv2.TM_CCOEFF&＃39;, &＃39;cv2.TM_CCOEFF_NORMED&＃39;, &＃39;cv2.TM_CCORR&＃39;,&＃39;cv2.TM_CCORR_NORMED&＃39;, &＃39;cv2.TM_SQDIFF&＃39;, &＃39;cv2.TM_SQDIFF_NORMED&＃39;]for meth in methods:img &＃61; img2.copy()method &＃61; eval(meth)# Apply template Matchingres &＃61; cv2.matchTemplate(img,template,method)min_val, max_val, min_loc, max_loc &＃61; cv2.minMaxLoc(res)# If the method is TM_SQDIFF or TM_SQDIFF_NORMED, take minimumif method in [cv2.TM_SQDIFF, cv2.TM_SQDIFF_NORMED]:top_left &＃61; min_locelse:top_left &＃61; max_locbottom_right &＃61; (top_left[0] &＃43; w, top_left[1] &＃43; h)cv2.rectangle(img,top_left, bottom_right, 255, 2)plt.subplot(121),plt.imshow(res,cmap &＃61; &＃39;gray&＃39;)plt.title(&＃39;Matching Result&＃39;), plt.xticks([]), plt.yticks([])plt.subplot(122),plt.imshow(img,cmap &＃61; &＃39;gray&＃39;)plt.title(&＃39;Detected Point&＃39;), plt.xticks([]), plt.yticks([])plt.suptitle(meth)plt.show()

与多个对象匹配的模板

cv.minMaxLoc()将不会提供所有的位置.在这种情况下&＃xff0c;我们将使用阈值.

import cv2
import numpy as np
from matplotlib import pyplot as pltimg_rgb &＃61; cv2.imread(&＃39;img5.png&＃39;)
img_gray &＃61; cv2.cvtColor(img_rgb, cv2.COLOR_BGR2GRAY)
template &＃61; cv2.imread(&＃39;img_roi1.png&＃39;,0)
w, h &＃61; template.shape[::-1]res &＃61; cv2.matchTemplate(img_gray,template,cv2.TM_CCOEFF_NORMED)
threshold &＃61; 0.8
loc &＃61; np.where( res >&＃61; threshold)
for pt in zip(*loc[::-1]):cv2.rectangle(img_rgb, pt, (pt[0] &＃43; w, pt[1] &＃43; h), (0,0,255), 2)cv2.imshow(&＃39;res&＃39;,img_rgb)