请给出一段完整代码，演示基于OpenCV的车牌识别

好的，以下是一个基于OpenCV的车牌识别的Python示例代码：

import cv2
import numpy as np

# 车牌定位
def locate_plate(img):
    # 转换为灰度图像
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    # 高斯模糊
    blur = cv2.GaussianBlur(gray, (5, 5), 0)
    # 边缘检测
    edges = cv2.Canny(blur, 100, 200)
    # 膨胀操作
    dilate_kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (5, 5))
    dilate = cv2.dilate(edges, dilate_kernel)
    # 轮廓检测
    contours, hierarchy = cv2.findContours(dilate, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
    # 筛选可能的车牌区域
    plate_contours = []
    for cnt in contours:
        area = cv2.contourArea(cnt)
        x, y, w, h = cv2.boundingRect(cnt)
        aspect_ratio = float(w) / h
        if area > 1000 and aspect_ratio > 2 and aspect_ratio < 5:
            plate_contours.append(cnt)
    # 绘制可能的车牌区域
    plate_img = img.copy()
    cv2.drawContours(plate_img, plate_contours, -1, (0, 255, 0), 2)
    return plate_img, plate_contours

# 车牌字符分割
def segment_chars(img, plate_contours):
    # 转换为灰度图像
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    # 二值化
    _, thresh = cv2.threshold(gray, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)
    # 获取车牌区域的坐标范围
    plate_rect = cv2.boundingRect(plate_contours[0])
    # 裁剪出车牌区域
    plate = thresh[plate_rect[1]:plate_rect[1] + plate_rect[3], plate_rect[0]:plate_rect[0] + plate_rect[2]]
    # 垂直投影
    vertical_sum = np.sum(plate, axis=0)
    # 寻找字符分割点
    index = []
    for i in range(len(vertical_sum)):
        if vertical_sum[i] > 0:
            index.append(i)
    index.append(plate.shape[1])
    segments = []
    for i in range(len(index) - 1):
        x1 = index[i]
        x2 = index[i + 1]
        segment = plate[:, x1:x2]
        segments.append(segment)
    # 绘制字符分割线
    plate_img = img.copy()
    for i in index:
        cv2.line(plate_img, (plate_rect[0] + i, plate_rect[1]), (plate_rect[0] + i, plate_rect[1] + plate_rect[3]), (0, 255, 0), 2)
    return plate_img, segments

# 车牌字符识别
def recognize_chars(segments):
    # 加载字符分类器
    svm = cv2.ml.SVM_load('svm.xml')
    # 字符识别
    chars = []
    for segment in segments:
        # 调整大小
        h, w = segment.shape
        if h > 0 and w > 0:
            if h > w:
                scale = 32 / h
                h = 32
                w = int(w * scale)
            else:
                scale = 32 / w
                w = 32
                h = int(h * scale)
            resized = cv2.resize(segment, (w, h), interpolation=cv2.INTER_AREA)
            # 填充
            pad_h = (64 - h) // 2
            pad_w = (64 - w) // 2
            padded = cv2.copyMakeBorder(resized, pad_h, pad_h, pad_w, pad_w, cv2.BORDER_CONSTANT, value=0)
            # 特征提取
            feature = np.reshape(padded, (1, 64 * 64))
            # 预测
            response = svm.predict(feature.astype(np.float32))[1]
            char = chr(response + ord('A'))
            chars.append(char)
    return chars

# 主函数
if __name__ == '__main__':
    # 读取图像
    img = cv2.imread('plate.jpg')
    # 车牌定位
    plate_img, plate_contours = locate_plate(img)
    # 车牌字符分割
    plate_img, segments = segment_chars(img, plate_contours)
    # 车牌字符识别
    chars = recognize_chars(segments)
    # 显示结果
    print(''.join(chars))
    cv2.imshow('Plate', plate_img)
    cv2.waitKey(0)

需要注意的是，以上代码仅供参考，实际应用中还需要根据具体情况进行适当调整和优化。