五行Python实现验证码识别，太稳了！(Python实战：五行代码搞定验证码识别，超稳技巧分享！)

原创

ithorizon 6个月前 (10-21) 阅读数 24 #后端开发

五行Python实现验证码识别，太稳了！

一、前言

在当今互联网时代，验证码是一种常见的网络平安手段，用于防止自动化脚本或恶意机器人进行大规模的注册、登录等操作。验证码识别技术也应运而生，成为了人工智能领域的一个热门应用。本文将介绍怎样使用Python五行代码实现验证码的识别，这种方法不仅简洁，而且精确率高，非常稳定。

二、验证码识别原理

验证码识别通常分为以下几个步骤：图像预处理、字符分割、字符识别。下面我们将详细解释这三个步骤。

三、图像预处理

图像预处理是识别验证码的第一步，核心包括图像灰度化、二值化、去噪、腐蚀和膨胀等操作。

3.1 图像灰度化

将彩色图像变成灰度图像，以简化后续处理。


import cv2
def convert_to_gray(image_path):
    image = cv2.imread(image_path)
    gray_image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    return gray_image

3.2 图像二值化

将灰度图像变成二值图像，即像素值只有0和255两种。


def binary_image(image):
    _, binary_image = cv2.threshold(image, 127, 255, cv2.THRESH_BINARY)
    return binary_image

3.3 噪声去除

去除图像中的噪声，减成本时间识别精确率。


def remove_noise(image):
    kernel = np.ones((3, 3), np.uint8)
    denoised_image = cv2.morphologyEx(image, cv2.MORPH_CLOSE, kernel)
    denoised_image = cv2.morphologyEx(denoised_image, cv2.MORPH_OPEN, kernel)
    return denoised_image

3.4 腐蚀和膨胀

腐蚀和膨胀操作可以去除小的噪点，使字符更加明确。


def erode_and_dilate(image):
    kernel = np.ones((3, 3), np.uint8)
    eroded_image = cv2.erode(image, kernel, iterations=1)
    dilated_image = cv2.dilate(eroded_image, kernel, iterations=1)
    return dilated_image

四、字符分割

字符分割是将图像中的每个字符分离出来，为后续的字符识别做准备。

4.1 查找字符轮廓

通过查找轮廓来定位每个字符。


def find_contours(image):
    contours, _ = cv2.findContours(image, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
    return contours

4.2 提取字符

按照轮廓信息提取每个字符。


def extract_characters(contours, image):
    characters = []
    for contour in contours:
        x, y, w, h = cv2.boundingRect(contour)
        character = image[y:y+h, x:x+w]
        characters.append(character)
    return characters

五、字符识别

字符识别是验证码识别的核心部分，可以使用深度学习、模板匹配等方法。

5.1 模板匹配

使用模板匹配对字符进行识别。


def template_matching(characters, templates):
    recognized_text = ""
    for character in characters:
        max_score = 0
        best_match = None
        for template_name, template in templates.items():
            score = cv2.matchTemplate(character, template, cv2.TM_CCOEFF_NORMED)
            if score > max_score:
                max_score = score
                best_match = template_name
        recognized_text += best_match
    return recognized_text

六、完整代码示例

下面是一个完整的验证码识别示例，包括图像预处理、字符分割和字符识别。


import cv2
import numpy as np
def convert_to_gray(image_path):
    image = cv2.imread(image_path)
    gray_image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    return gray_image
def binary_image(image):
    _, binary_image = cv2.threshold(image, 127, 255, cv2.THRESH_BINARY)
    return binary_image
def remove_noise(image):
    kernel = np.ones((3, 3), np.uint8)
    denoised_image = cv2.morphologyEx(image, cv2.MORPH_CLOSE, kernel)
    denoised_image = cv2.morphologyEx(denoised_image, cv2.MORPH_OPEN, kernel)
    return denoised_image
def erode_and_dilate(image):
    kernel = np.ones((3, 3), np.uint8)
    eroded_image = cv2.erode(image, kernel, iterations=1)
    dilated_image = cv2.dilate(eroded_image, kernel, iterations=1)
    return dilated_image
def find_contours(image):
    contours, _ = cv2.findContours(image, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
    return contours
def extract_characters(contours, image):
    characters = []
    for contour in contours:
        x, y, w, h = cv2.boundingRect(contour)
        character = image[y:y+h, x:x+w]
        characters.append(character)
    return characters
def template_matching(characters, templates):
    recognized_text = ""
    for character in characters:
        max_score = 0
        best_match = None
        for template_name, template in templates.items():
            score = cv2.matchTemplate(character, template, cv2.TM_CCOEFF_NORMED)
            if score > max_score:
                max_score = score
                best_match = template_name
        recognized_text += best_match
    return recognized_text
# 加载模板
templates = {
    '0': cv2.imread('template_0.png', 0),
    '1': cv2.imread('template_1.png', 0),
    '2': cv2.imread('template_2.png', 0),
    '3': cv2.imread('template_3.png', 0),
    '4': cv2.imread('template_4.png', 0),
    '5': cv2.imread('template_5.png', 0),
    '6': cv2.imread('template_6.png', 0),
    '7': cv2.imread('template_7.png', 0),
    '8': cv2.imread('template_8.png', 0),
    '9': cv2.imread('template_9.png', 0)
}
# 读取验证码图像
image_path = 'captcha.jpg'
gray_image = convert_to_gray(image_path)
binary_image = binary_image(gray_image)
denoised_image = remove_noise(binary_image)
eroded_image = erode_and_dilate(denoised_image)
# 分割字符
contours = find_contours(eroded_image)
characters = extract_characters(contours, eroded_image)
# 识别字符
recognized_text = template_matching(characters, templates)
print('识别导致：', recognized_text)