基于SVM的验证码识别

__author__ = 'Directorli'

import Image

import os

import svmlight

def binary(x,y,file):

    im = Image.open(file)

    Lim = im.convert('L')

    threshold = 80

    table = []

    for i in range(256):

        if i <>

            table.append(0)

        else:

            table.append(1)

    # convert to binary image by the table

    bim = Lim.point(table, '1')

    bdata=bim.load()

    out=''

    for a in range(y):

        for b in range(x):

            out = out + str(bdata[b,a])

        #out = out + '\n'

    return out

def totrain(x):

    path = 'bmp\\cut\\all\\'

    filelist = os.listdir(path)

    #print filelist

    f=[]

    for one in filelist:

        # print path[-5]

        # print binary(10,10,path)

        a = binary(10,10,path+one)

        d=[]

        c = 1

        for b in a:

            d.append((c,int(b)))

            c = c + 1

        # print d

        if str(x) == one[-5]:

            e = (1,d)

        else:

            e = (-1,d)

        f.append(e)

    #print e

    #print f

    return f

def trainall():

    for i in range(10):

        training_data = totrain(i) #train set

        #test_data =test       #test set

        # train a model based on the data

        model = svmlight.learn(training_data, type='classification', verbosity=0)

        svmlight.write_model(model, 'model'+str(i))  #write model

        #model=svmlight.read_model('my_model.dat')    #read model

        #predictions = svmlight.classify(model, test_data)

生成的知识模式（model 0~9）：

 

7、再获取一定数量的验证码图片试着识别：
因为图片太规整，识别率太高（100%） ，所以对图片手动加些处理

以下是机器命名的，识别率任然算高

  
上代码： 
#ocr.py
#python ver 2.7

__author__ = 'Directorli'

import Image

import svmlight

def binary(x,file):

    im = Image.open(file)

    Lim = im.convert('L')

    threshold = 80

    table = []

    for i in range(256):

        if i <>

            table.append(0)

        else:

            table.append(1)

    # convert to binary image by the table

    bim = Lim.point(table, '1')

    bdata=bim.load()

    out=''

    for a in range(10):

        for b in range(x*10,x*10+10):

            out = out + str(bdata[b,a])

        #out = out + '\n'

    return out

def chformat(x):

    f=[]

    d=[]

    c = 1

    for b in x:

        d.append((c,int(b)))

        c = c + 1

    e = (1,d)

    f.append(e)

    return f

def ocr(filename):

    result=''

    for num in range(4):

        t =  binary(num,filename)

        test = chformat(t)

        for i in range(10):

            model=svmlight.read_model('model'+str(i))    #read model

            prediction = svmlight.classify(model,test)

            #print prediction

            if prediction[0] >0:

                #print i

                result = result + str(i)

    print(result)

    return result

#ocr('code.bmp')

import os

allfile = os.listdir('test\\')

for a in allfile:

    b = str(ocr('test\\'+a))