验证码（CAPTCHA，全自动区分计算机和人类的图灵测试）的缩写，是一种区分用户是计算机还是人工智能的全自动程序。
实验步骤：
1、创建验证码2、对验证码进行01值化3、降噪4、对验证码进行切分5、对切分后的验证码进行图片转数字化6、使用逻辑回归建模7、对新输入的图片进行预测
验证码的创建
1、随机生成验证码的颜色
2、随机生成验证码数字
3、使用PIL进行画图
import osfrom PIL import Imagefrom PIL import ImageDrawfrom PIL import ImageFontimport randomimport matplotlib.pyplot as pltdef getRandomColor():"""获取一个随机颜色(r,g,b)格式的:return:"""c1 = random.randint(0, 255)c2 = random.randint(0, 255)c3 = random.randint(0, 255)if c1 == 255:c1 = 0if c2 == 255:c2 = 0if c3 == 255:c3 = 0return (c1, c2, c3)def getRandomStr():"""获取一个随机数字，每个数字的颜色也是随机的:return:"""random_num = str(random.randint(0, 9))return random_numdef generate_captcha():"""使用PIL画图步骤:return: """# 获取一个Image对象，参数分别是RGB模式。宽150，高30， 随机颜色image = Image.new('RGB', (150, 50), (255, 255, 255))# 获取一个画笔对象，将图片对象传过去draw = ImageDraw.Draw(image)# 获取一个font字体对象参数是ttf的字体文件的目录，以及字体的大小font = ImageFont.truetype("arlrdbd.ttf", size=32) # 如果找不到字体，需要从网上下载到本地label = ""# 随机生成有5个数字的字符串for i in range(5):random_char = getRandomStr()label += random_char# 在图片上写东西，参数是：定位，字符串，颜色，字体draw.text((10+i*30, 0), random_char, getRandomColor(), font=font)# 画出随机噪点噪线width = 150height = 30# 画线for i in range(3):x1 = random.randint(0, width)x2 = random.randint(0, width)y1 = random.randint(0, height)y2 = random.randint(0, height)draw.line((x1, y1, x2, y2), fill=(0, 0, 0))# 画点for i in range(5):draw.point([random.randint(0, width), random.randint(0, height)], fill=getRandomColor())x = random.randint(0, width)y = random.randint(0, height)draw.arc((x, y, x + 4, y + 4), 0, 90, fill=(0, 0, 0))# 保存到硬盘，名为test.png格式为png的图片image.save(open(''.join(['captcha_images/', label, '.png']), 'wb'), 'png')# image.save(open(''.join(['captcha_predict/', label, '.png']), 'wb'), 'png')
执行代码之后，会在‘captcha_images’下生成实验所需的图片，如图：
图像处理：对生成的图片进行处理
(1)对验证码图片二值化，首先把图像从RGB 三通道转化成Gray单通道，然后把灰度图（0～255）转化成二值图（0，1）。
(2)将处理好的二值图进行降噪，去除图片中的噪点和噪线
from PIL import Imageimport numpy as npimport matplotlib.pyplot as pltimport osdef binarization(path):"""把一个rgb的图转换成一个二值图:param path::return:"""# 通过path把图像laod进来img = Image.open(path)# 把图像转化成一个灰度图img_gray = img.convert("L")# 把灰度图组装成数组形式img_gray = np.array(img_gray)# print(img_gray)# 得到灰度图的宽和高w, h = img_gray.shapefor x in range(w):for y in range(h):# 得到每一个像素块里的灰度值gray = img_gray[x, y]# 如果灰度值小于等于220， 就把它变成黑色if gray <= 220:img_gray[x, y] = 0# 如果灰度值大于220，就把它变成白色else:img_gray[x, y] = 1plt.figure("")plt.imshow(img_gray, cmap="gray")plt.axis("off")plt.show()return img_graydef noiseReduction(img_gray, label):"""降噪，也就是处理离群点如果一个像素点周围只有小于4个黑点的时候，那么这个点就是离群点:param img_gray::param label::return:"""height, width = img_gray.shapefor x in range(height):for y in range(width):cnt = 0# 白色的点不用管if img_gray[x, y] == 1:continueelse:try:if img_gray[x-1, y-1] == 0:cnt += 1except:passtry:if img_gray[x-1, y] == 0:cnt += 1except:passtry:if img_gray[x-1, y+1] == 0:cnt += 1except:passtry:if img_gray[x, y-1] == 0:cnt += 1except:passtry:if img_gray[x, y+1] == 0:cnt += 1except:passtry:if img_gray[x+1, y-1] == 0:cnt += 1except:passtry:if img_gray[x+1, y] == 0:cnt += 1except:passtry:if img_gray[x+1, y+1] == 0:cnt += 1except:passif cnt < 4: # 周围少于4个点就算是噪点img_gray[x, y] = 1plt.figure(" ")plt.imshow(img_gray, cmap="gray")plt.axis("off")plt.savefig("".join(["clean_captcha_img/", label, ".png"]))def image_2_clean():"""把所有的图像都转化成二值图:return:"""captchas = os.listdir("".join(["captcha_images/"]))for captcha in captchas:label = captcha.split(".")[0]image_path = "".join(["captcha_images/", captcha])# 二值化im = binarization(image_path)# 降噪noiseReduction(im, label)if __name__ == '__main__':image_2_clean()# path = "captcha_images/00006.png"# img_gray = binarization(path)# noiseReduction(img_gray, label='00006')
图像分割：对降噪后的图片进行分割，并对分割后的图片进行存储
import osfrom PIL import Imagefrom PIL import ImageDrawfrom PIL import ImageFontimport randomimport matplotlib.pyplot as pltdef cutImg(label):"""把图像的每一个数字都切分出来，并且存到新的文件夹下:param label::return:"""labels = list(label)img = Image.open("".join(['clean_captcha_img/', label, '.png']))for i in range(5):pic = img.crop((100*(1+i), 170, 100*(1+i)+100, 280))plt.imshow(pic)# seq就是我们需要存到文件的文件名seq = get_save_seq(label[i])pic.save("".join(["cut_number/", str(label[i]), "/", str(seq), '.png']))def get_save_seq(num):"""得到需要保存的数据的文件名每一个数文件下的文件名，都是从0开始保存 0.png, 1.png....:param num::return:"""nmlist = os.listdir("".join(["cut_number/", num, "/"]))if len(nmlist) == 0 or nmlist is None:return 0else:max_file = 0for file in nmlist:if int(file.split(".")[0]) > max_file:max_file = int(file.split(".")[0])return int(max_file) + 1def clean_to_cut():"""对每一个文件都进行切分:return:"""captchas = os.listdir("".join(["clean_captcha_img"]))for captcha in captchas:label = captcha.split(".")[0]cutImg(label)def create_dir():for i in range(10):os.mkdir("".join(["cut_number/", str(i)]))if __name__ == '__main__':# create_dir()clean_to_cut()
图片转数字化：对切分后的图片灰度化、二值化，使用Image.open()打开图片文件，得到plt图片对象，将plt图片对象转换为ndarray对象，将二值化后的图像转化为1行n列，存入X列表中，并将其对应的数字存入Y列表中。
模型的生成：将X，Y传入逻辑回归模型中，使用交叉验证和网格搜索寻找最优的参数。
import osfrom PIL import Imageimport numpy as npimport matplotlib.pyplot as pltfrom sklearn.linear_model import LogisticRegressionfrom sklearn.model_selection import train_test_splitfrom sklearn.model_selection import GridSearchCVfrom sklearn.externals import joblibfrom sklearn.metrics import confusion_matrixfrom sklearn.metrics import precision_scorefrom sklearn.metrics import recall_scoredef load_data():"""把数据从cut_number里面导出来其中X指的是每一个数字的01值的排列， Y指的是这个数字是什么:return:"""X, Y = [], []cut_list = os.listdir("cut_number")# 循环cut_number文件夹下的每一个自文件夹（1，2，3，4，5...）for numC in cut_list:num_list_dir = "".join(["cut_number/", str(numC), "/"])nums_dir = os.listdir(num_list_dir)# 循环子文件夹中的每一个图片# print(np.array(Image.open(''.join(['cut_number/', str(numC), '/', '0.png']))))for num_file in nums_dir:# 导入数字图片img = Image.open("".join(["cut_number/", str(numC), "/", num_file]))# print(np.array(img))# 对数字图片做灰度化img_gray = img.convert("L")# plt.imshow(img_gray)# 把灰度化图片保存到数组里img_array = np.array(img_gray)w, h = img_array.shape# 把灰度化的图片做二值化for x in range(w):for y in range(h):gray = img_array[x, y]if gray <= 220:img_array[x, y] = 0else:img_array[x, y] = 1# 把二值化的图片reshape成1行，n列img_re = img_array.reshape(1, -1)# print(img_re[0])X.append(img_re[0])Y.append(int(numC))return np.array(X), np.array(Y)def generate_model(X, Y):"""生成模型:param X::param Y::return:"""# 区分测试集和训练集，37开X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=0.3)log_clf = LogisticRegression(multi_class="ovr", solver="sag", max_iter=10000)# log_clf.fit(X_train, Y_train)# 利用交叉验证选择参数param_grid = {"tol": [1e-4, 1e-5, 1e-2], "C": [0.4, 0.6, 0.8]}grid_search = GridSearchCV(log_clf, param_grid=param_grid, cv=3)grid_search.fit(X, Y)print(grid_search.best_params_)print("模型生成成功")# 将模型持久化joblib.dump(log_clf, "captcha_model/captcha_model.model")print("模型保存成功")if __name__ == '__main__':X, Y = load_data()generate_model(X, Y)
图片的预测：
输入要预测的图片，对其进行灰度化，二值化，并进行分割，将分割出来的五个图片输入进模型中。
import osfrom PIL import Imageimport numpy as npimport matplotlib.pyplot as pltfrom sklearn.linear_model import LogisticRegressionfrom sklearn.model_selection import train_test_splitfrom sklearn.model_selection import GridSearchCVfrom sklearn.externals import joblibfrom sklearn.metrics import confusion_matrixfrom sklearn.metrics import precision_scorefrom sklearn.metrics import recall_scorefrom .captcha_logistic import *def get_model():model = joblib.load('captcha_model/captcha_model.model')return modeldef model_predict():path = 'captcha_predict/unknown.png'pre_img_gray = binarization(path)noiseReduction(pre_img_gray, 'unknown')# cut imagelabels = ['0', '1', '2', '3', '4']img = Image.open(''.join(['clean_captcha_img/unknown.png']))for i in range(5):pic = img.crop((100 * (1 + i), 170, 100 * (1 + i) + 100, 280))plt.imshow(pic)pic.save(''.join(['captcha_predict/', labels[i], '.png']))result = ''model = get_model()for i in range(5):path = ''.join(['captcha_predict/', labels[i], '.png'])img = Image.open(path)img_gray = img.convert('L')img_array = np.array(img_gray)w, h = img_array.shapefor x in range(w):for y in range(h):gray = img_array[x, y]if gray <= 220:img_array[x, y] = 0else:img_array[x, y] = 1img_re = img_array.reshape(1, -1)X = img_re[0]y_pre = model.predict([X])result = ''.join([result, str(y_pre[0])])return resultif __name__ == '__main__':result = model_predict()print(result)
