from cv2 import waitKey
|
import numpy
|
import pickle
|
import cv2
|
import os
|
import uuid
|
|
#图片预处理
|
def pretreatment(image):
|
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
|
thresh = cv2.threshold(gray, 0, 255,cv2.THRESH_BINARY_INV | cv2.THRESH_OTSU)[1]
|
return thresh
|
|
#提取图片特征
|
'''
|
def feature(A):
|
midx=int(A.shape[1]/2)+1
|
midy=int(A.shape[0]/2)+1
|
A1=cv2.mean(A[0:midy,0:midx])
|
A2=cv2.mean(A[midy:A.shape[0],0:midx])
|
A3=cv2.mean(A[0:midy,midx:A.shape[1]])
|
A4=cv2.mean(A[midy:A.shape[0],midx:A.shape[1]])
|
A5=cv2.mean(A)
|
AF=[A1,A2,A3,A4,A5]
|
return AF
|
'''
|
def feature(A):
|
midx=int(A.shape[1]/2)+1
|
midy=int(A.shape[0]/2)+1
|
A1=cv2.mean(A[0:int(A.shape[0]/5),0:A.shape[1]])
|
A2=cv2.mean(A[0:midy,0:int(A.shape[1]/3)])
|
A3=cv2.mean(A[0:midy,int(A.shape[1]*2/3):A.shape[1]])
|
A4=cv2.mean(A[int(A.shape[0]*3/5):int(A.shape[0]*4/5),0:A.shape[1]])
|
A5=cv2.mean(A[midy:A.shape[0],0:int(A.shape[1]/3)])
|
A6=cv2.mean(A[midy:A.shape[0],int(A.shape[1]*2/3):A.shape[1]])
|
A7=cv2.mean(A[int(A.shape[0]*4/5):A.shape[0],0:A.shape[1]])
|
if A1[0] >128 :
|
A1=1
|
else:
|
A1=0
|
if A2[0] >128 :
|
A2=1
|
else:
|
A2=0
|
if A3[0] >128 :
|
A3=1
|
else:
|
A3=0
|
if A4[0] >128 :
|
A4=1
|
else:
|
A4=0
|
if A5[0] >128 :
|
A5=1
|
else:
|
A5=0
|
if A6[0] >128 :
|
A6=1
|
else:
|
A6=0
|
if A7[0] >128 :
|
A7=1
|
else:
|
A7=0
|
|
AF=[A1,A2,A3,A4,A5,A6,A7]
|
#print(AF)
|
i=0
|
tube=0
|
onenumber=0
|
while(i < 7):
|
if(AF[i]==1):
|
tube = tube + pow(2,i)
|
i +=1
|
if(tube==63):
|
onenumber = 0
|
elif(tube==6):
|
onenumber = 1
|
elif(tube==91):
|
onenumber = 2
|
elif(tube==79):
|
onenumber = 3
|
elif(tube==102 or tube==110):
|
#110是因为有干扰情况
|
onenumber = 4
|
elif(tube==109):
|
onenumber = 5
|
elif(tube==125):
|
onenumber = 6
|
elif(tube==7):
|
onenumber = 7
|
elif(tube==127):
|
onenumber = 8
|
elif(tube==103):
|
onenumber = 9
|
else:
|
onenumber = -1
|
|
return AF,onenumber
|
|
|
|
#训练已知图片的特征
|
def training():
|
train_set={}
|
for i in range(11):
|
value=[]
|
for j in range(10):
|
#print('./'+str(i)+'/'+str(i)+'-'+str(j)+'.jpg')
|
ima=cv2.imread('./data/'+str(i)+'.jpg',0)
|
AF,nmber = feature(ima)
|
value.append(AF)
|
#print(value)
|
train_set[i]=value
|
#把训练结果存为永久文件，以备下次使用
|
output=open('./train_set.pkl','wb')
|
pickle.dump(train_set,output)
|
output.close()
|
return train_set
|
|
#计算两向量的距离
|
def distance(v1,v2):
|
vector1=numpy.array(v1)
|
vector2=numpy.array(v2)
|
Vector=(vector1-vector2)**2
|
distance = Vector.sum()**0.5
|
return distance
|
|
#用最近邻算法识别单个数字
|
def knn(train_set,V,k):
|
key_sort=[11]*k
|
value_sort=[11]*k
|
for key in range(11):
|
for value in train_set[key]:
|
d=distance(V,value)
|
for i in range(k):#从小到大排序
|
if d<value_sort[i]:
|
for j in range(k-2,i-1,-1):
|
key_sort[j+1]=key_sort[j]
|
value_sort[j+1]=value_sort[j]
|
key_sort[i]=key
|
value_sort[i]=d
|
break
|
max_key_count=-1
|
key_set=set(key_sort)
|
for key in key_set: #统计每个类别出现的次数
|
if max_key_count<key_sort.count(key):
|
max_key_count=key_sort.count(key)
|
max_key=key
|
return max_key
|
|
#合并一副图片的所有数字
|
def identification(train_set,AF,k):
|
key=knn(train_set,AF,k)
|
return key
|
|
#识别文件夹中的所有图片上的数据并输出到EXCELL中
|
def main(k,trained=None):
|
# k：knn算法的的k，即取训练集中最相邻前k个样本进行判别
|
#trained：trained=0则程序会开始训练出训练集，否则会用已保存的训练集
|
if trained==0:
|
train_set=training()
|
else:
|
pkl_file=open('train_set.pkl','rb')
|
train_set=pickle.load(pkl_file)
|
pkl_file.close()
|
#识别所有图片的数字并输出到EXCELL中
|
path = "./images4"
|
i=0
|
filename = os.listdir(path) # 获取文件夹文件
|
for file in filename:
|
img = cv2.imread(path + "/" + file,0) # 读取图片
|
AF,nmber=feature(img)
|
print(AF,nmber)
|
result=identification(train_set,AF,k)
|
cv2.imwrite('./images4_infer/'+str(result)+'_'+str(uuid.uuid4())+".jpg",img)
|
i = i+ 1
|
if i >200 :
|
break
|
|
main(1,0) #表示：4近邻，还没有训练集备份，还没有昨日数据备份
|
