python实现连连看辅助（图像识别）
个人兴趣，用python实现连连看的辅助程序，总结实现过程及知识点。
总体思路
1、获取连连看程序的窗口并前置
2、游戏界面截图，将每个一小图标切图，并形成由小图标组成的二维列表
3、对图片的二维列表遍历，将二维列表转换成由数字组成的二维数组，图片相同的数值相同。
4、遍历二维数组，找到可消除的对象，实现算法：
实现过程中遇到的问题
图片切割
im = image.crop((left,top,right,bottom))
//image.crop参数为一个列表或元组，顺序为(left,top,right,bottom)
找到游戏运行窗口
hdwd = win32gui.FindWindow(0,wdname)
# 设置为最前显示
win32gui.SetForegroundWindow(hdwd)
窗口不要点击最小化，点击后无法弹出来。
图片缩放并转为灰度img1 = im1.resize((20, 20), Image.ANTIALIAS).convert('L')
Image.ANTIALIAS 为抗锯齿的选项，图片无毛边。
获取图片每个点的RGB值pi1 = list(img1.getdata())
列表每个元素为一个三位数的值，分别代表该点的RGB值。列表pi1共400个元素。（因为图片为20*20）
鼠标点击消除PyMouse.click()该方法默认双击，改为PyMouse.press() 或 PyMouse.release()
判断图片相似 汉明距离，平均哈希
def compare_img(self,im1,im2):
img1 = im1.resize((20, 20), Image.ANTIALIAS).convert('L')
img2 = im2.resize((20, 20), Image.ANTIALIAS).convert('L')
pi1 = list(img1.getdata())
pi2 = list(img2.getdata())
avg1 = sum(pi1) / len(pi1)
avg2 = sum(pi2) / len(pi2)
hash1 = "".join(map(lambda p: "1" if p > avg1 else "0", pi1))
hash2 = "".join(map(lambda p: "1" if p > avg2 else "0", pi2))
match = 0
for i in range(len(hash1)):
if hash1[i] != hash2[i]:
match += 1
# match = sum(map(operator.ne, hash1, hash2))
# match 值越小，相似度越高
return match
计算直方图
from PIL import Image
# 将图片转化为RGB
def make_regalur_image(img, size=(8, 8)):
gray_image = img.resize(size).convert('RGB')
return gray_image
# 计算直方图
def hist_similar(lh, rh):
assert len(lh) == len(rh)
hist = sum(1 - (0 if l == r else float(abs(l - r)) / max(l, r)) for l, r in zip(lh, rh)) / len(lh)
return hist
# 计算相似度
def calc_similar(li, ri):
calc_sim = hist_similar(li.histogram(), ri.histogram())
return calc_sim
if __name__ == '__main__':
image1 = Image.open('1-10.jpg')
image1 = make_regalur_image(image1)
image2 = Image.open('2-11.jpg')
image2 = make_regalur_image(image2)
print("图片间的相似度为", calc_similar(image1, image2))
# 值在[0,1]之间，数值越大，相似度越高
图片余弦相似度
from PIL import Image
from numpy import average, dot, linalg
# 对图片进行统一化处理
def get_thum(image, size=(64, 64), greyscale=False):
# 利用image对图像大小重新设置, Image.ANTIALIAS为高质量的
image = image.resize(size, Image.ANTIALIAS)
if greyscale:
# 将图片转换为L模式，其为灰度图，其每个像素用8个bit表示
image = image.convert('L')
return image
# 计算图片的余弦距离
def image_similarity_vectors_via_numpy(image1, image2):
image1 = get_thum(image1)
image2 = get_thum(image2)
images = [image1, image2]
vectors = []
norms = []
for image in images:
vector = []
for pixel_tuple in image.getdata():
vector.append(average(pixel_tuple))
vectors.append(vector)
# linalg=linear（线性）+algebra（代数），norm则表示范数
# 求图片的范数？？
norms.append(linalg.norm(vector, 2))
a, b = vectors
a_norm, b_norm = norms
# dot返回的是点积，对二维数组（矩阵）进行计算
res = dot(a / a_norm, b / b_norm)
return res
if __name__ == '__main__':
image1 = Image.open('1-9.jpg')
image2 = Image.open('8-6.jpg')
cosin = image_similarity_vectors_via_numpy(image1, image2)
print('图片余弦相似度', cosin)
# 值在[0,1]之间，数值越大，相似度越高,计算量较大，效率较低
完整代码
import win32gui
import time
from PIL import ImageGrab , Image
import numpy as np
from pymouse import PyMouse
class GameAuxiliaries(object):
def __init__(self):
self.wdname = r'宠物连连看经典版2,宠物连连看经典版2小游戏,4399小游戏 www.4399.com - Google Chrome'
# self.wdname = r'main.swf - PotPlayer'
self.image_list = {}
self.m = PyMouse()
def find_game_wd(self,wdname):
# 取得窗口句柄
hdwd = win32gui.FindWindow(0,wdname)
# 设置为最前显示
win32gui.SetForegroundWindow(hdwd)
time.sleep(1)
def get_img(self):
image = ImageGrab.grab((417, 289, 884, 600))
# image = ImageGrab.grab((417, 257, 885, 569))
image.save('1.jpg','JPEG')
for x in range(1,9):
self.image_list[x] = {}
for y in range(1,13):
top = (x - 1) * 38 + (x-2)
left =(y - 1) * 38 +(y-2)
right = y * 38 + (y-1)
bottom = x * 38 +(x -1)
if top < 0:
top = 0
if left < 0 :
left = 0
im_temp = image.crop((left,top,right,bottom))
im = im_temp.crop((1,1,37,37))
im.save('{}-{}.jpg'.format(x,y))
self.image_list[x][y]=im
# 判断两个图片是否相同。汉明距离，平均哈希
def compare_img(self,im1,im2):
img1 = im1.resize((20, 20), Image.ANTIALIAS).convert('L')
img2 = im2.resize((20, 20), Image.ANTIALIAS).convert('L')
pi1 = list(img1.getdata())
pi2 = list(img2.getdata())
avg1 = sum(pi1) / len(pi1)
avg2 = sum(pi2) / len(pi2)
hash1 = "".join(map(lambda p: "1" if p > avg1 else "0", pi1))
hash2 = "".join(map(lambda p: "1" if p > avg2 else "0", pi2))
match = 0
for i in range(len(hash1)):
if hash1[i] != hash2[i]:
match += 1
# match = sum(map(operator.ne, hash1, hash2))
# match 值越小，相似度越高
return match
# 将图片矩阵转换成数字矩阵
def create_array(self):
array = np.zeros((10,14),dtype=np.int32)
img_type_list = []
for row in range(1,len(self.image_list)+1):
for col in range(1,len(self.image_list[1])+1):
# im = Image.open('{}-{}.jpg'.format(row,col))
im = self.image_list[row][col]
for img in img_type_list:
match = self.compare_img(im,img)
# match = test2.image_similarity_vectors_via_numpy(im,img)
if match <15:
array[row][col] = img_type_list.index(img) +1
break
else:
img_type_list.append(im)
array[row][col] = len(img_type_list)
return array
def row_zero(self,x1,y1,x2,y2,array):
'''相同的图片中间图标全为空'''
if x1 == x2:
min_y = min(y1,y2)
max_y = max(y1,y2)
if max_y - min_y == 1:
return True
for y in range(min_y+1,max_y):
if array[x1][y] != 0 :
return False
return True
else:
return False
def col_zero(self,x1,y1,x2,y2,array):
'''相同的图片同列'''
if y1 == y2:
min_x = min(x1,x2)
max_x = max(x1,x2)
if max_x - min_x == 1:
return True
for x in range(min_x+1,max_x):
if array[x][y1] != 0 :
return False
return True
else:
return False
def two_line(self,x1,y1,x2,y2,array):
'''两条线相连，转弯一次'''
for row in range(1,9):
for col in range(1,13):
if row == x1 and col == y2 and array[row][col]==0 and self.row_zero(x1,y1,row,col,array) and self.col_zero(x2,y2,row,col,array):
return True
if row == x2 and col == y1 and array[row][col]==0 and self.row_zero(x2,y2,row,col,array) and self.col_zero(x1,y1,row,col,array):
return True
return False
def three_line(self,x1,y1,x2,y2,array):
'''三条线相连，转弯两次'''
for row1 in range(10):
for col1 in range(14):
for row2 in range(10):
for col2 in range(14):
if array[row1][col1] == array[row2][col2] == 0 and self.row_zero(x1,y1,row1,col1,array) and self.row_zero(x2,y2,row2,col2,array) and self.col_zero(row1,col1,row2,col2,array):
return True
if array[row1][col1] == array[row2][col2] == 0 and self.col_zero(x1,y1,row1,col1,array) and self.col_zero(x2,y2,row2,col2,array) and self.row_zero(row1,col1,row2,col2,array):
return True
if array[row1][col1] == array[row2][col2] == 0 and self.row_zero(x2,y2,row1,col1,array) and self.row_zero(x1,y1,row2,col2,array) and self.col_zero(row1,col1,row2,col2,array):
return True
if array[row1][col1] == array[row2][col2] == 0 and self.col_zero(x2,y2,row1,col1,array) and self.col_zero(x1,y1,row2,col2,array) and self.row_zero(row1,col1,row2,col2,array):
return True
return False
def mouse_click(self,x,y):
top = (x - 1) * 38 + (x - 2)
left = (y - 1) * 38 + (y - 2)
right = y * 38 + (y - 1)
bottom = x * 38 + (x - 1)
if top < 0:
top = 0
if left < 0:
left = 0
self.m.press(int(417+(left+right)/2) ,int(289+(top+bottom)/2) )
def find_same_img(self,array):
for x1 in range(1,9):
for y1 in range(1,13):
if array[x1][y1] == 0:
continue
for x2 in range(1,9):
for y2 in range(1,13):
if x1==x2 and y1 == y2:
continue
if array[x2][y2] == 0 :
continue
if array[x1][y1] != array[x2][y2] :
continue
if array[x1][y1] ==array[x2][y2] and (self.row_zero(x1,y1,x2,y2,array) or self.col_zero(x1,y1,x2,y2,array) or self.two_line(x1,y1,x2,y2,array) or self.three_line(x1,y1,x2,y2,array)):
print("可消除！x{}y{} 和 x{}y{}".format(x1,y1,x2,y2))
self.mouse_click(x1,y1)
time.sleep(0.1)
self.mouse_click(x2,y2)
time.sleep(0.1)
array[x1][y1]=array[x2][y2]=0
def run(self):
#找到游戏运行窗口
self.find_game_wd(self.wdname)
# 截图，切割成小图标
self.get_img()
# 将图片矩阵转换成数字矩阵
array = self.create_array()
print(array)
# 遍历矩阵，找到可消除项，点击消除
for i in range(10):
self.find_same_img(array)
print(array)
if __name__ == '__main__':
ga = GameAuxiliaries()
ga.run()
总结
该程序其实未能完全实现辅助功能，主要是因为图片切割时未找到更好的规则，造成图片识别困难，缩放比例和判断阀值未找到一个平衡点，阀值太大，则将不同的图标识别为相同，阀值太小，相同的图标又判断为不一样。
您可能感兴趣的文章:python实现连连看辅助之图像识别延伸Python 连连看连接算法python递归法实现简易连连看小游戏
