基于OpenCV的车牌识别

本文将从各种角度介绍基于OpenCV的车牌识别，包括车牌定位、车牌字符分割及识别等方面。

一、车牌定位

车牌定位是车牌识别的第一步，其目的是在整张图像中找出车牌的位置，并将车牌区域从图像中提取出来。车牌定位可以采用基于颜色的方法或基于形状的方法。

以下是基于颜色的车牌定位代码示例：

import cv2
import numpy as np

def locate_car_plate(img):
    # 转换为HSV空间
    hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
    # 设定蓝色区域的HSV范围
    lower_blue = np.array([100, 50, 50])
    upper_blue = np.array([124, 255, 255])
    # 根据HSV范围生成掩膜
    mask = cv2.inRange(hsv, lower_blue, upper_blue)
    # 对掩膜进行开闭运算
    kernel = np.ones((5, 5), np.uint8)
    mask = cv2.morphologyEx(mask, cv2.MORPH_OPEN, kernel)
    mask = cv2.morphologyEx(mask, cv2.MORPH_CLOSE, kernel)
    # 找出轮廓
    contours, hierarchy = cv2.findContours(mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
    # 找出最大轮廓
    max_area = 0
    max_contour = None
    for contour in contours:
        area = cv2.contourArea(contour)
        if area > max_area:
            max_area = area
            max_contour = contour
    # 找出轮廓的外接矩形
    x, y, w, h = cv2.boundingRect(max_contour)
    # 在原图像上绘制外接矩形
    cv2.rectangle(img, (x, y), (x + w, y + h), (0, 255, 0), 2)
    # 返回车牌图像
    plate_img = img[y:y + h, x:x + w]
    return plate_img

二、车牌字符分割

车牌字符分割的目的是将车牌中的字符区域提取出来进行识别。字符分割可以采用基于形状或基于颜色的方法。

以下是基于形状的车牌字符分割代码示例：

def segment_car_plate_characters(img):
    # 转换为灰度图像
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    # 进行OTSU阈值分割
    ret, binary = cv2.threshold(gray, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
    # 找出轮廓
    contours, hierarchy = cv2.findContours(binary, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
    # 对轮廓进行排序
    contours = sorted(contours, key=lambda c: cv2.boundingRect(c)[0])
    # 提取字符图像
    character_imgs = []
    for contour in contours:
        x, y, w, h = cv2.boundingRect(contour)
        if w > h:
            character_img = img[y:y + h, x:x + w]
            character_img = cv2.resize(character_img, (20, 20))
            character_imgs.append(character_img)
    return character_imgs

三、车牌字符识别

车牌字符识别的目的是将字符图像中的字符进行识别。字符识别可以采用基于SVM或基于深度学习的方法。

以下是基于SVM的车牌字符识别代码示例：

import joblib

def recognize_car_plate_characters(character_imgs):
    # 加载分类器模型
    svm = joblib.load('svm.pkl')
    # 提取特征
    features = []
    for img in character_imgs:
        gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
        feature = hog(gray, orientations=9, pixels_per_cell=(8, 8), cells_per_block=(3, 3))
        features.append(feature)
    # 进行分类
    pred_labels = svm.predict(features)
    # 转换为字符
    characters = []
    for label in pred_labels:
        characters.append(chr(label))
    return characters

四、实例演示

本文提供了一个基于OpenCV的简单车牌识别实例，可以通过以下代码进行测试：

import cv2
import numpy as np
import joblib
from skimage.feature import hog

def locate_car_plate(img):
    hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
    lower_blue = np.array([100, 50, 50])
    upper_blue = np.array([124, 255, 255])
    mask = cv2.inRange(hsv, lower_blue, upper_blue)
    kernel = np.ones((5, 5), np.uint8)
    mask = cv2.morphologyEx(mask, cv2.MORPH_OPEN, kernel)
    mask = cv2.morphologyEx(mask, cv2.MORPH_CLOSE, kernel)
    contours, hierarchy = cv2.findContours(mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
    max_area = 0
    max_contour = None
    for contour in contours:
        area = cv2.contourArea(contour)
        if area > max_area:
            max_area = area
            max_contour = contour
    x, y, w, h = cv2.boundingRect(max_contour)
    cv2.rectangle(img, (x, y), (x + w, y + h), (0, 255, 0), 2)
    plate_img = img[y:y + h, x:x + w]
    return plate_img

def segment_car_plate_characters(img):
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    ret, binary = cv2.threshold(gray, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
    contours, hierarchy = cv2.findContours(binary, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
    contours = sorted(contours, key=lambda c: cv2.boundingRect(c)[0])
    character_imgs = []
    for contour in contours:
        x, y, w, h = cv2.boundingRect(contour)
        if w > h:
            character_img = img[y:y + h, x:x + w]
            character_img = cv2.resize(character_img, (20, 20))
            character_imgs.append(character_img)
    return character_imgs

def recognize_car_plate_characters(character_imgs):
    svm = joblib.load('svm.pkl')
    features = []
    for img in character_imgs:
        gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
        feature = hog(gray, orientations=9, pixels_per_cell=(8, 8), cells_per_block=(3, 3))
        features.append(feature)
    pred_labels = svm.predict(features)
    characters = []
    for label in pred_labels:
        characters.append(chr(label))
    return characters

if __name__ == "__main__":
    img = cv2.imread("test.jpg")
    plate_img = locate_car_plate(img)
    character_imgs = segment_car_plate_characters(plate_img)
    characters = recognize_car_plate_characters(character_imgs)
    print(characters)

结论

本文从车牌定位、车牌字符分割及识别等方面详细介绍了基于OpenCV的车牌识别。通过实例演示可以发现该方法在车牌定位和字符分割方面表现较好，但字符识别需要更多的训练数据和优化才能达到更好的效果。