Image in Image with cvMatchTemplate - but how?

Question

I want to find out at which position of a source-image a certain sub-image appears (e.g. source image: http://i.pictr.com/6xg895m69q.png, sub-image: http://i.pictr.com/jdaz9zwzej.png). As far as I know it is necessary to transform the arrays to make them "readable" to OpenCV, this is what I tried, but for some reason, it does not work. here is my code so far:

 from PIL import Image
 import numpy
 from pylab import *
 import cv2
 import cv

 image = cv2.imread('source_img.jpg')
 template = cv2.imread('template_img.jpg')

 im = cv.fromarray(image)
 templ = cv.fromarray(template)
 result = numpy.zeros(shape=(1,10)) ##create a matrix with 0s
 a = cv.fromarray(result)
 cv.MatchTemplate(im, templ, a, cv.CV_TM_CCORR)
 print result
 print image

my goal is to write the coordinates of the sub-images in the result array (the rest of the array should keep the value 0 (I know that my code wont make this up to now). This the error message, I get when executing the code:

OpenCV Error: Assertion failed (result.size() == cv::Size(std::abs(img.cols - templ.cols) + 1, std::abs(img.rows - templ.rows) + 1) && result.type() == CV_32F) in cvMatchTemplate, file /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_graphics_opencv/opencv/work/OpenCV-2.4.3/modules/imgproc/src/templmatch.cpp, line 376 Traceback (most recent call last): File "/Users/strongbow/imagerecognition.py", line 27, in cv.MatchTemplate(im, templ, a, cv.CV_TM_CCORR) cv2.error: result.size() == cv::Size(std::abs(img.cols - templ.cols) + 1, std::abs(img.rows - templ.rows) + 1) && result.type() == CV_32F

I am new to OpenCV and really don't know what to do with this error-message. Anyone an idea/pointer what to do?

Answer 1

import sys
import cv2
import numpy

img = cv2.imread(sys.argv[1])
template = cv2.imread(sys.argv[2])
th, tw = template.shape[:2]

result = cv2.matchTemplate(img, template, cv2.TM_CCORR_NORMED)
threshold = 0.99
loc = numpy.where(result >= threshold)
for pt in zip(*loc[::-1]):
    cv2.rectangle(img, pt, (pt[0] + tw, pt[1] + th), 0, 2)

cv2.imwrite(sys.argv[3], img)

Answer 2

import cv2
from cv2 import cv

image = cv2.imread('1_tree.jpg')
template = cv2.imread('1_tree_detail.jpg')

values = cv2.matchTemplate(image, template, method=cv.CV_TM_SQDIFF)
best_fit_point = cv2.minMaxLoc(values)[2]
bottom_right = best_fit_point[0]+template.shape[0], best_fit_point[1]+template.shape[1]
cv2.rectangle(image, best_fit_point, bottom_right, (255,255,255))
cv2.imshow('tree',image)
cv2.imwrite('tree_match.jpg', image)
cv2.waitKey()