要想運行該代碼，請確保安裝了：python 2.7，opencv 2.4.9

效果如下：

算法如下：

把圖片先進行處理，處理過程：

1.用膨脹圖像與腐蝕圖像相減的方法獲得輪廓。

2.用二值化獲得圖像

3. 反色

經過如上的處理之后，圖片為：

這之后就簡單了，設計一個辦法把三種圖像區分開來即可。

代碼如下：

# -*- coding: cp936 -*-
import cv2
import numpy
import time
import random
import os
def judge( ):
    #構造一個3×3的結構元素
    # return 0 stone ,1 jiandao, 2 bu
    img = cv2.imread("wif.jpg",0)
    element = cv2.getStructuringElement(cv2.MORPH_RECT,(11,11))
    dilate = cv2.dilate(img, element)
    erode = cv2.erode(img, element)
    #將兩幅圖像相減獲得邊，第一個參數是膨脹后的圖像，第二個參數是腐蝕后的圖像
    result = cv2.absdiff(dilate,erode);
    #上面得到的結果是灰度圖，將其二值化以便更清楚的觀察結果
    retval, result = cv2.threshold(result, 40, 255, cv2.THRESH_BINARY);

    #反色，即對二值圖每個像素取反
    result = cv2.bitwise_not(result);
    result =cv2.medianBlur(result,23)
    a=[]
    posi =[]
    width =[]
    count = 0
    area = 0 
    for i in range(result.shape[1]):
        for j in range(result.shape[0]):
            if(result[j][i]==0):
                area+=1
    for i in range(result.shape[1]):
        if(result[5*result.shape[0]/16][i]==0 and result[5*result.shape[0]/16][i-1]!=0 ):
            count+=1
            width.append(0)
            posi.append(i)
        if(result[5*result.shape[0]/16][i]==0):
            width[count-1]+=1
    """
    print 'the pic width is ',result.shape[1],'
'
    for i in range(count):
        print 'the ',i,'th',' ','is';
        print 'width ',width[i]
        print 'posi ',posi[i],'
'
    print count,'
'
    print 'area is ',area,'
'
    
    cv2.line(result,(0,5*result.shape[0]/16),(214,5*result.shape[0]/16),(0,0,0))
    cv2.namedWindow("fcuk")
    cv2.imshow("fcuk",result)
    cv2.waitKey(0)
    """
    #判定時間
    
    width_length=0
    width_jiandao = True
    for i in range(count):
        if width[i]>45:
            #print 'bu1';
            return 2;
        if width[i]<=20 or width[i]>=40:
            width_jiandao= False
        width_length += width[i]
    if width_jiandao==True and count==2:
        return 1;
    if(area <8500):
        #print 'shi tou';
        return 0;
    print "width_leng",width_length
    if(width_length<35):
        #這個時候說明照片是偏下的，所以需要重新測定。
        a=[]
        posi =[]
        width =[]
        count = 0
        for i in range(result.shape[1]):
            if(result[11*result.shape[0]/16][i]==0 and result[11*result.shape[0]/16][i-1]!=0 ):
                count+=1
                width.append(0)
                posi.append(i)
            if(result[11*result.shape[0]/16][i]==0):
                width[count-1]+=1
        """
        print 'the pic width is ',result.shape[1],'
'
        for i in range(count):
            print 'the ',i,'th',' ','is';
            print 'width ',width[i]
            print 'posi ',posi[i],'
'
        print count,'
'
        print 'area is ',area,'
'
        """
    width_length=0
    width_jiandao = True
    for i in range(count):
        if width[i]>45:
            #print 'bu1';
            return 2;
        if width[i]<=20 or width[i]>=40:
            width_jiandao= False
        width_length += width[i]
    if width_jiandao==True and count==2:
        return 1;
    if(area>14000 or count>=3):
        #print 'bu2';
        return 2;
    if(width_length<110):
        #print 'jian dao';
        return 1;
    else:
        #print 'bu3';
        return 2;
 
 
"""
print("這是通過攝像頭來玩的剪刀石頭布的游戲，輸入y開始
")
s = raw_input()
capture = cv2.VideoCapture(0)
cv2.namedWindow("camera",1)
start_time = time.time()
print("給你5秒的時間把手放到方框的位置
")
while(s=='y' or s=='Y'):
    ha,img =capture.read()
    end_time = time.time()
    cv2.rectangle(img,(426,0),(640,250),(170,170,0))
    cv2.putText(img,str(int((5-(end_time- start_time)))), (100,100), cv2.FONT_HERSHEY_SIMPLEX, 2, 255)
    cv2.imshow("camera",img)
    
    if(end_time-start_time>5):
        break
    if(cv2.waitKey(30)>=0):
        break
ha,img = capture.read()
capture.release()
cv2.imshow("camera",img)
img = img[0:210,426:640]
cv2.imwrite("wif.jpg",img)
judge()  
cv2.waitKey(0)
print "fuck"
"""
def game():
    fuck =[]
    fuck.append("石頭")
    fuck.append("剪刀")
    fuck.append("布")
    capture = cv2.VideoCapture(0)
    cv2.namedWindow("camera",1)
    start_time = time.time()
    print("給你5秒的時間把手放到方框的位置
")
    while(1):
        ha,img =capture.read()
        end_time = time.time()
        cv2.rectangle(img,(426,0),(640,250),(170,170,0))
        cv2.putText(img,str(int((5-(end_time- start_time)))), (100,100), cv2.FONT_HERSHEY_SIMPLEX, 2, 255)
        cv2.imshow("camera",img)
        if(end_time-start_time>5):
            break
        if(cv2.waitKey(30)>=0):
            break
    ha,img = capture.read()
    capture.release()
    cv2.imshow("camera",img)
    img = img[0:210,426:640]
    cv2.imwrite("wif.jpg",img)
    p1 = judge()
    pc = random.randint(0,2)
    #print p1,' ',pc,'
'
    print "你出的是",fuck[p1]," 電腦出的是",fuck[pc],"
"
    cv2.destroyAllWindows()
    if(p1==pc):
        print "平局
"
        return 0
    if((p1==0 and pc ==1)or(p1==1 and pc ==2)or(p1==2 and pc ==0)):
        print '你贏了
'
        return 1
    else:
        print '你輸了
'
        return -1
def main():
    you_win=0
    pc_win=0
    print("這是通過攝像頭來玩的剪刀石頭布的游戲，請輸入回車開始游戲
")
    s = raw_input()
    while(1):
        print "比分(玩家：電腦) ",you_win,":",pc_win,'
'
        s = raw_input()
        os.system('cls')
        ans =game()
        if(ans == 1):
            you_win+=1
        elif(ans == -1):
            pc_win+=1
        print "為了減少誤判，請盡可能將手占據盡可能大的框框"
main()