在图像内部查找图像（使用 CV2）

Question

I am using CV2 and I want to detect an image inside of an image. Here is the situation:

I have this base image, and I am trying to detect the current character that is displayed. There are around 30 characters in the game, so I was thinking about making a png for every character (character1.png, character2.png, etc.) so I can find the current character the user is playing. Here is an example of the character1.png template:

I want to match the template to that a region of the image. The problem is, if multiple people in the game are playing that same character, their characters will get detected as well. But, I only want the client's character to get detected. The client's character will always be on the bottom left quadrant of the game.

This is my code:

import cv2
import numpy as np

img_bgr = cv2.imread('base.png')
img_gry = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2GRAY)

template = cv2.imread('search.png', 0)

w, h = template.shape[::-1]

res = cv2.matchTemplate(img_gry, template, cv2.TM_CCOEFF_NORMED)
threshold = 0.8
loc = np.where(res >= threshold)

for pt in zip(*loc[::-1]):
    cv2.rectangle(img_bgr, pt, (pt[0]+w, pt[1]+h), (0,255,255), 2)

cv2.imshow('detected', img_bgr)

So far, it functions slightly how I want it to. It puts the yellow rectangle around the character:

However, if there were to be multiple people with the same character, the other people's character would get detected as well. I was wondering if there exists a feature in cv2 to only search within a certain region of the base image (that bottom left region where the client's character is). Moreover, I do not need the yellow rectangle to be displayed, that was only as a test. So, I was wondering if I could just make cv2 say "character 1 detected", "character 2 detected", etc if the template was found in the region.

So, basically, I want my program to cycle through characters(1-30).png, and as soon as it finds the correct character the client is playing, it will say "You are playing character N (n=1-30). I am wondering if this is an efficient way of detecting the client's character.

Answer 1

You can select ROI (region of interest) of the area where the character is. And then check if the template is found only in this area:

import cv2
import numpy as np

img_bgr = cv2.imread('base.png')
img_gry = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2GRAY)

template = cv2.imread('search.png', 0)

w, h = template.shape[::-1]
roi_start = (430, 550) # top left corner of the selected area
roi_end = (526,651) # bottom right corner of the selected area

roi = img_gry[roi_start[1]: roi_end[1], roi_start[0]: roi_end[0]]
res = cv2.matchTemplate(roi, template, cv2.TM_CCOEFF_NORMED)
threshold = 0.8
loc = np.where(res >= threshold)
for pt in zip(*loc[::-1]):
    if pt is not None:
        print("character found")
        break