先放制作好的游戏视频链接:(纯粹是兴趣分享)
连连看不一样的玩法-图像相似度识别-python_单机游戏热门视频
https://www.ixigua.com/7076826558106698253?logTag=f2d215ea6e6372c45362
这两个视频是我本人自制投稿,感兴趣可以观看一下。下面是截图展示:
程序主要功能是先将练练看的整个大图切分成单个小图,然后进行循环遍历找出相似的图片,并在矩阵中进行记录,然后依据练练看两个图片连接的规则进行连接。直接放代码:
1.相似度计算模块(用一个算法就行了,我这里写了好几个)
import cv2
import numpy as np
import matplotlib
matplotlib.use('TkAgg')
均值哈希算法
def aHash(img):
# 缩放为8*8
# img = cv2.imread(img)
img = np.asanyarray(img)
img = cv2.resize(img, (8, 8))
# 转换为灰度图
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# s为像素和初值为0,hash_str为hash值初值为''
s = 0
hash_str = ''
# 遍历累加求像素和
for i in range(8):
for j in range(8):
s = s+gray[i, j]
# 求平均灰度
avg = s/64
# 灰度大于平均值为1相反为0生成图片的hash值
for i in range(8):
for j in range(8):
if gray[i, j] > avg:
hash_str = hash_str+'1'
else:
hash_str = hash_str+'0'
return hash_str
差值哈希算法
def dHash(img):
# 缩放8*8
# img = cv2.imread(img) #此种方法只能用于本地图片读取
img = np.asanyarray(img)
img = cv2.resize(img, (9, 8))
# 转换灰度图
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
hash_str = ''
# 每行前一个像素大于后一个像素为1,相反为0,生成哈希
for i in range(8):
for j in range(8):
if gray[i, j] > gray[i, j+1]:
hash_str = hash_str+'1'
else:
hash_str = hash_str+'0'
return hash_str
感知哈希算法
def pHash(img):
# 缩放32*32
# img = cv2.imread(img)
img = np.asanyarray(img)
img = cv2.resize(img, (32, 32)) # , interpolation=cv2.INTER_CUBIC
# 转换为灰度图
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# 将灰度图转为浮点型,再进行dct变换
dct = cv2.dct(np.float32(gray))
# opencv实现的掩码操作
dct_roi = dct[0:8, 0:8]
hash = []
avreage = np.mean(dct_roi)
for i in range(dct_roi.shape[0]):
for j in range(dct_roi.shape[1]):
if dct_roi[i, j] > avreage:
hash.append(1)
else:
hash.append(0)
return hash
灰度直方图算法
def calculate(image1, image2):
# 计算单通道的直方图的相似值
# 均值、差值、感知哈希算法三种算法值越小,则越相似,相同图片值为0
# 三直方图算法和单通道的直方图 0-1之间,值越大,越相似。 相同图片为1
# image1 = cv2.imread(image1)
# image2 = cv2.imread(image2)
image1 = np.asanyarray(image1)
image2 = np.asanyarray(image2)
hist1 = cv2.calcHist([image1], [0], None, [256], [0.0, 255.0])
hist2 = cv2.calcHist([image2], [0], None, [256], [0.0, 255.0])
# 计算直方图的重合度
degree = 0
for i in range(len(hist1)):
if hist1[i] != hist2[i]:
degree = degree + \
(1 - abs(hist1[i] - hist2[i]) / max(hist1[i], hist2[i]))
else:
degree = degree + 1
degree = degree / len(hist1)
return degree
三通道直方图算法
def classify_calculate(image1, image2):
# 计算单通道的直方图的相似值
# 均值、差值、感知哈希算法三种算法值越小,则越相似,相同图片值为0
# 三直方图算法和单通道的直方图 0-1之间,值越大,越相似。 相同图片为1
hist1 = cv2.calcHist([image1], [0], None, [256], [0.0, 255.0])
hist2 = cv2.calcHist([image2], [0], None, [256], [0.0, 255.0])
# 计算直方图的重合度
degree = 0
for i in range(len(hist1)):
if hist1[i] != hist2[i]:
degree = degree + \
(1 - abs(hist1[i] - hist2[i]) / max(hist1[i], hist2[i]))
else:
degree = degree + 1
degree = degree / len(hist1)
return degree
RGB每个通道的直方图相似度
def classify_hist_with_split(image1, image2, size=(256, 256)):
# 将图像resize后,分离为RGB三个通道,再计算每个通道的相似值
image1 = np.asanyarray(image1)
image2 = np.asanyarray(image2)
image1 = cv2.resize(image1, size)
image2 = cv2.resize(image2, size)
sub_image1 = cv2.split(image1)
sub_image2 = cv2.split(image2)
sub_data = 0
for im1, im2 in zip(sub_image1, sub_image2):
sub_data += classify_calculate(im1, im2)
sub_data = sub_data / 3
return sub_data
Hash值对比
def cmpHash(hash1, hash2):
# 均值、差值、感知哈希算法三种算法值越小,则越相似,相同图片值为0
# 三直方图算法和单通道的直方图 0-1之间,值越大,越相似。 相同图片为1
# 算法中1和0顺序组合起来的即是图片的指纹hash。顺序不固定,但是比较的时候必须是相同的顺序。
# 对比两幅图的指纹,计算汉明距离,即两个64位的hash值有多少是不一样的,不同的位数越小,图片越相似
# 汉明距离:一组二进制数据变成另一组数据所需要的步骤,可以衡量两图的差异,汉明距离越小,则相似度越高。汉明距离为0,即两张图片完全一样
n = 0
# hash长度不同则返回-1代表传参出错
if len(hash1) != len(hash2):
return -1
# 遍历判断
for i in range(len(hash1)):
# 不相等则n计数+1,n最终为相似度
if hash1[i] != hash2[i]:
n = n + 1
return n
2.基本功能(截图,图片转换)
from PIL import ImageGrab
import numpy as np
#截图方法
#传入图片像素位置以及所需比例宽度高度 #引入自己写的模块
from LLKgame.similarFunctinos import cmpHash, pHash, classify_hist_with_split, calculate
#传入截图的左上右下坐标,宽的小图数量,长的小图数量,小图的高度和宽度
#如cutPicture(304, 271, 1226, 870,10,14,60.2,65.9)在坐标(304,271)(1226,870)处截取140个小图,小图的高是60.2,宽是65.9
def cutPicture(left,top,right,botton,pWidth,pHeight,height,width): # 截图
size = (left, top, right, botton)
img = ImageGrab.grab(size)
# img.size(200,100)
print(img.size)
print(img)
# img.save("D://cut.jpg")
# img.show()
# 2、分切小图
image_list = {}
for x in range(pWidth):
image_list[x] = {}
for y in range(pHeight):
# print("show",x, y)
# exit()
top1 = x * height
left1 = y * width
right1 = (y + 1) * width
botton1 = (x + 1) * height
# 用crop函数切割成小图标,参数为图标的左上角和右下角左边
im = img.crop((left1, top1, right1, botton1))
# im.show()
# time.sleep(1)
# 将切割好的图标存入对应的位置
image_list[x][y] = im
print(image_list)
return image_list
#创造数字矩阵
#传入矩阵的宽高和初始矩阵
def makeArray(pWidth,pHeight,image_list):
image_type_list = []
arr = np.zeros((pWidth + 2,pHeight + 2), dtype=np.int32) # 创建矩阵以数字代替图片
for i in range(len(image_list)):
for j in range(len(image_list[0])):
im = image_list[i][j]
# 验证当前图标是否已存入
index = getIndex(10,0.65,im, image_type_list)
# 不存在image_type_list
if index < 0:
image_type_list.append(im)
arr[i + 1][j + 1] = len(image_type_list)
else:
arr[i + 1][j + 1] = index + 1
print("图标数:", len(image_type_list))
# self.im2num_arr = arr
return arr
检查数组中是否有图标,如果有则返回索引下表
#传入标准相似度similar1 similar2 和需要验证图片
def getIndex(similar1,similar2,im, im_list):
global flag
for i in range(len(im_list)):
flag = 0
# if self.compare_image_with_hash(im, im_list[i],6):
# return i
val1 = calculate(im,im_list[i])
val2 = classify_hist_with_split(im,im_list[i])
val3 = cmpHash(pHash(im), pHash(im_list[i]))
if val2 >= similar2:
flag = 1
if val1 >= similar2:
flag = flag + 1
if val3 <= 1 similar1: flag="flag" + if>= 2:
return i
return -1
</=>
3.连连看连接逻辑
import time
from pymouse import *
from LLKgame.baseFunctions import cutPicture, makeArray
class run_game():
#初始截图左上右下坐标位置,初始截图总宽总高,初始单个图像宽高
def __init__(self,left,top,right,botton,pWidth,pHeight,height,width):
self.left = left
self.top = top
self.right = right
self.botton = botton
self.pWidth = pWidth
self.pHeight = pHeight
self.width = width
self.height = height
self.im2num_arr = []
#点击事件
def pClick(self, x1, y1, x2, y2):
m = PyMouse()
p1_x = int(self.left + (y1) * self.width - int((self.width / 2)))
p1_y = int(self.top + (x1) * self.height - int((self.height / 2)))
p2_x = int(self.left + (y2) * self.width - int((self.width / 2)))
p2_y = int(self.top + (x2) * self.height - int((self.height / 2)))
time.sleep(0.3)#沉睡时间
m.click(int(p1_x / 1.25), int(p1_y / 1.25))
time.sleep(0.3)
m.click(int(p2_x / 1.25), int(p2_y / 1.25))
# time.sleep(0.1)
# 设置矩阵值为0
self.im2num_arr[x1][y1] = 0
self.im2num_arr[x2][y2] = 0
print("消除:(%d, %d) (%d, %d)" % (x1, y1, x2, y2))
# 是否为同行或同列且可连
#X1 Y1 元素坐标x值以及y值 X2 Y2 元素坐标x值 y值
def isReachable(self, x1, y1, x2, y2):
# 1、先判断值是否相同
if self.im2num_arr[x1][y1] != self.im2num_arr[x2][y2]:
return False
# 判断横向连通
if self.isSameRow(x1, y1, x2, y2):
return True
# 判断纵向连通
if self.isSameCol(x1, y1, x2, y2):
return True
# 判断一个拐点可连通
if self.turnOnceCheck(x1, y1, x2, y2):
return True
# 判断两个拐点可连通
if self.turnTwiceCheck(x1, y1, x2, y2):
return True
# 不可联通返回False
return False
#是否两个元素同行
def isSameRow(self, x1, y1, x2, y2):
if (abs(y1 - y2) == 1 and x1 == x2): # 同行且相邻
return True
if (abs(y1 - y2) > 1 and x1 == x2): # 同行不相邻
# if (x1 == 0 or x1 == (self.pHeight - 1)): # 最外行
# return True
flag = 0
for i in range(min(y1, y2) + 1, max(y1, y2)):
if self.im2num_arr[x1][i] == 0:
flag = flag + 0
else:
flag = flag + 1
if (flag == 0):
return True
return False
#是否同列
def isSameCol(self, x1, y1, x2, y2):
if (abs(x1 - x2) == 1 and y1 == y2): # 同列且相邻
return True
if (abs(x1 - x2) > 1 and y1 == y2): # 同列不相邻
# if (y1 == 0 or y1 == (self.pWidth - 1)): # 在最外列
# return True
flag = 0
for i in range(min(x1, x2) + 1, max(x1, x2)):
if self.im2num_arr[i][y1] == 0:
flag = flag + 0
else:
flag = flag + 1
if (flag == 0):
return True
return False
# 判断一个拐点可联通
def turnOnceCheck(self, x1, y1, x2, y2):
if x1 == x2 or y1 == y2:
return False
cx = x1
cy = y2
dx = x2
dy = y1
# 拐点为空,从第一个点到拐点并且从拐点到第二个点可通,则整条路可通。
if self.im2num_arr[cx][cy] == 0:
if self.isSameRow(x1, y1, cx, cy) and self.isSameCol(cx, cy, x2, y2):
return True
if self.im2num_arr[dx][dy] == 0:
if self.isSameRow(x1, y1, dx, dy) and self.isSameCol(dx, dy, x2, y2):
return True
return False
# 判断两个拐点可联通
def turnTwiceCheck(self, x1, y1, x2, y2):
if x1 == x2 and y1 == y2:
return False
# 遍历整个数组找合适的拐点
for i in range(0, len(self.im2num_arr)):
for j in range(0, len(self.im2num_arr[1])):
# 不为空不能作为拐点
if self.im2num_arr[i][j] != 0:
continue
# 不和被选方块在同一行列的不能作为拐点
if i != x1 and i != x2 and j != y1 and j != y2:
continue
# 作为交点的方块不能作为拐点
if (i == x1 and j == y2) or (i == x2 and j == y1):
continue
if self.turnOnceCheck(x1, y1, i, j) and (
self.isSameRow(i, j, x2, y2) or self.isSameCol(i, j, x2, y2)):
return True
if self.turnOnceCheck(i, j, x2, y2) and (
self.isSameRow(x1, y1, i, j) or self.isSameCol(x1, y1, i, j)):
return True
return False
# 判断矩阵是否全为0
def isAllZero(self, arr):
for i in range(0, self.pWidth + 2):
for j in range(0, self.pHeight + 2):
if arr[i][j] != 0:
return False
return True
def start(self):
# 3、遍历查找可以相连的坐标
global num1
global num2
num1 = 0#循环判断
num2 = 0 #结束符
print(self.im2num_arr)
while not self.isAllZero(self.im2num_arr):
if num1 == 10:
for i in range(0, self.pWidth + 2):
for j in range(0, self.pHeight + 2):
self.im2num_arr[i][j] = 0
num2 = 1
for x1 in range(0, self.pWidth + 2):
for y1 in range(0, self.pHeight + 2):
if self.im2num_arr[x1][y1] == 0:
continue
for x2 in range(0, self.pWidth + 2):
for y2 in range(0, self.pHeight + 2):
if self.im2num_arr[x2][y2] == 0 or (x1 == x2 and y1 == y2):
continue
if self.im2num_arr[x1][y1] == self.im2num_arr[x2][y2]:
if self.isReachable(x1, y1, x2, y2):
self.pClick(x1, y1, x2, y2)
num1 = num1 + 1
return num2
if __name__ == '__main__':
t = run_game(304, 271, 1226, 870,10,14,60.2,65.9)#设置元素坐标
time.sleep(3)#等待3秒
image = cutPicture(304, 271, 1226, 870,10,14,60.2,65.9)#截图
arr = makeArray(10,14,image)#设置矩阵
t.im2num_arr = arr
num = t.start()
print("游戏结束!")
总结:程序难点在于两个图片之间相似度识别的精确性,不过对于练练看这个游戏来说,这个程序所使用的计算算法是够用的。
遇到问题:程序往往错误判断两个图片可以连接
解决方案:细微调整截图以及小图切分的像素位置,往往小图的切分是问题导致的关键,只要调整好切分的像素坐标,是没有问题的。
再次放视频链接:感兴趣可以看看,这是自制文章和视频,纯粹是兴趣分享,请在视频中点个赞吧!https://www.ixigua.com/7076826558106698253?logTag=f2d215ea6e6372c45362
说明:文章中引用了部分网友的知识点,不过这个程序我写的太久远的,不知道是哪篇文章了,如有引用请在评论区说明,我加个链接。
Original: https://blog.csdn.net/cqq1171422470/article/details/123965283
Author: QQ的猫好懒
Title: 连连看不一样的玩法,利用python进行图片相似度计算
原创文章受到原创版权保护。转载请注明出处:https://www.johngo689.com/642004/
转载文章受原作者版权保护。转载请注明原作者出处!