341 lines
11 KiB
Python
341 lines
11 KiB
Python
import cv2 as cv
|
|
import numpy as np
|
|
import matplotlib.pyplot as plt
|
|
import argparse
|
|
import typing
|
|
import logging
|
|
from pathlib import Path
|
|
from dataclasses import dataclass
|
|
|
|
# Reference:
|
|
# - Claude Code
|
|
# - https://www.cnblogs.com/linuxAndMcu/p/19144795
|
|
# - https://www.51halcon.com/forum.php?mod=viewthread&tid=6562
|
|
|
|
UNI_HW: int = 1000
|
|
|
|
|
|
def _uniform_car_plate(img: cv.typing.MatLike) -> cv.typing.MatLike:
|
|
"""
|
|
Uniform the image size to 512x512 while maintaining aspect ratio, padding with black.
|
|
|
|
:param img: The image in BGR format to be uniformed.
|
|
:return: The uniformed image in BGR format.
|
|
"""
|
|
# Calculate the new width and height for given image
|
|
h, w = img.shape[:2]
|
|
scale = min(UNI_HW / w, UNI_HW / h)
|
|
new_w = int(w * scale)
|
|
new_h = int(h * scale)
|
|
|
|
# Resize the image
|
|
resized_img = cv.resize(img, (new_w, new_h))
|
|
|
|
# Create a black canvas of size UNI_HW x UNI_HW
|
|
padded_img = np.zeros((UNI_HW, UNI_HW, 3), dtype=np.uint8)
|
|
|
|
# Calculate position to paste the resized image (centered)
|
|
y_offset = (UNI_HW - new_h) // 2
|
|
x_offset = (UNI_HW - new_w) // 2
|
|
|
|
# Paste the resized image onto the canvas
|
|
padded_img[y_offset : y_offset + new_h, x_offset : x_offset + new_w] = resized_img
|
|
|
|
# Return the padded image
|
|
return padded_img
|
|
|
|
|
|
@dataclass
|
|
class CarPlateHsvBoundary:
|
|
lower_bound: cv.typing.MatLike
|
|
upper_bound: cv.typing.MatLike
|
|
|
|
|
|
CAR_PLATE_HSV_BOUNDARIES: tuple[CarPlateHsvBoundary, ...] = (
|
|
# 中国蓝牌 HSV 范围
|
|
CarPlateHsvBoundary(np.array([100, 80, 60]), np.array([130, 255, 255])),
|
|
# 中国绿牌 HSV 范围
|
|
CarPlateHsvBoundary(np.array([35, 43, 46]), np.array([99, 255, 255])),
|
|
# 中国黄牌 HSV 范围
|
|
CarPlateHsvBoundary(np.array([32, 43, 46]), np.array([68, 255, 255])),
|
|
)
|
|
|
|
|
|
def _batchly_mask_car_plate(
|
|
hsv: cv.typing.MatLike,
|
|
) -> typing.Iterator[cv.typing.MatLike]:
|
|
""" """
|
|
for boundary in CAR_PLATE_HSV_BOUNDARIES:
|
|
# 以给定HSV范围检测符合该颜色的位置
|
|
mask = cv.inRange(hsv, boundary.lower_bound, boundary.upper_bound)
|
|
|
|
# 形态学：闭运算填孔 + 开运算去噪
|
|
kernel_close = cv.getStructuringElement(cv.MORPH_RECT, (25, 10))
|
|
kernel_open = cv.getStructuringElement(cv.MORPH_RECT, (5, 5))
|
|
mask = cv.morphologyEx(mask, cv.MORPH_CLOSE, kernel_close)
|
|
mask = cv.morphologyEx(mask, cv.MORPH_OPEN, kernel_open)
|
|
|
|
# Return value
|
|
yield mask
|
|
|
|
|
|
@dataclass
|
|
class CarPlateRegion:
|
|
x: int
|
|
y: int
|
|
w: int
|
|
h: int
|
|
|
|
|
|
MIN_AREA: float = 3000
|
|
MIN_ASPECT_RATIO: float = 1.5
|
|
MAX_ASPECT_RATIO: float = 6.0
|
|
BEST_ASPECT_RATIO: float = 3.5
|
|
|
|
|
|
def _analyse_car_plate_connection(
|
|
mask: cv.typing.MatLike,
|
|
) -> typing.Optional[CarPlateRegion]:
|
|
# 连通域分析，筛选最符合车牌长宽比的区域
|
|
num_labels, labels, stats, _ = cv.connectedComponentsWithStats(mask, connectivity=8)
|
|
|
|
best: typing.Optional[CarPlateRegion] = None
|
|
best_score = 0
|
|
|
|
for i in range(1, num_labels):
|
|
x, y, w, h, area = stats[i]
|
|
# 检查面积
|
|
if area < MIN_AREA:
|
|
continue
|
|
# 标准车牌宽高比约 3:1 ~ 5:1
|
|
ratio = w / (h + 1e-5)
|
|
if ratio >= MIN_ASPECT_RATIO and ratio <= MAX_ASPECT_RATIO:
|
|
score = area * (1 - abs(ratio - BEST_ASPECT_RATIO) / BEST_ASPECT_RATIO)
|
|
if score > best_score:
|
|
best_score = score
|
|
best = CarPlateRegion(x, y, w, h)
|
|
|
|
return best
|
|
|
|
|
|
@dataclass
|
|
class PerspectiveData:
|
|
top_left: tuple[int, int]
|
|
top_right: tuple[int, int]
|
|
bottom_left: tuple[int, int]
|
|
bottom_right: tuple[int, int]
|
|
|
|
new_width: int
|
|
new_height: int
|
|
|
|
|
|
def _extract_perspective_data(
|
|
gray: cv.typing.MatLike,
|
|
) -> typing.Optional[PerspectiveData]:
|
|
""" """
|
|
# Histogram balance to increase contrast
|
|
hist_gray = cv.equalizeHist(gray)
|
|
|
|
# Apply Gaussian blur to reduce noise
|
|
blurred = cv.GaussianBlur(hist_gray, (5, 5), 0)
|
|
|
|
# Edge detection using Canny
|
|
edges = cv.Canny(blurred, 50, 150)
|
|
|
|
# Find contours
|
|
contours, _ = cv.findContours(edges, cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE)
|
|
if not contours:
|
|
return None
|
|
# Find the largest one because all image is car plate
|
|
max_area_contour = max(contours, key=lambda contour: cv.contourArea(contour))
|
|
|
|
# Approximate the contour
|
|
peri = cv.arcLength(max_area_contour, True)
|
|
approx = cv.approxPolyDP(max_area_contour, 0.02 * peri, True)
|
|
if len(approx) != 4:
|
|
return None
|
|
|
|
# Perspective transformation to get front view
|
|
# Order points: top-left, top-right, bottom-right, bottom-left
|
|
pts = approx.reshape(4, 2)
|
|
rect = np.zeros((4, 2), dtype="float32")
|
|
|
|
# Sum and difference of coordinates to find corners
|
|
s = pts.sum(axis=1)
|
|
top_left = pts[np.argmin(s)] # Top-left has smallest sum
|
|
bottom_right = pts[np.argmax(s)] # Bottom-right has largest sum
|
|
|
|
diff = np.diff(pts, axis=1)
|
|
top_right = pts[np.argmin(diff)] # Top-right has smallest difference
|
|
bottom_left = pts[np.argmax(diff)] # Bottom-left has largest difference
|
|
|
|
# Calculate width and height of new image
|
|
width_a = np.linalg.norm(rect[0] - rect[1])
|
|
width_b = np.linalg.norm(rect[2] - rect[3])
|
|
max_width = max(int(width_a), int(width_b))
|
|
|
|
height_a = np.linalg.norm(rect[0] - rect[3])
|
|
height_b = np.linalg.norm(rect[1] - rect[2])
|
|
max_height = max(int(height_a), int(height_b))
|
|
|
|
# Return value
|
|
return PerspectiveData(
|
|
top_left, top_right, bottom_left, bottom_right, max_width, max_height
|
|
)
|
|
|
|
|
|
def extract_car_plate(img: cv.typing.MatLike) -> typing.Optional[cv.typing.MatLike]:
|
|
"""
|
|
Extract the car plate part from given image.
|
|
|
|
:param img: The image containing car plate in BGR format.
|
|
:return: The image of binary car plate in U8 format if succeed, otherwise None.
|
|
"""
|
|
img = _uniform_car_plate(img)
|
|
|
|
# 转换到HSV空间
|
|
hsv = cv.cvtColor(img, cv.COLOR_BGR2HSV)
|
|
|
|
# 利用车牌颜色在 HSV 空间定位车牌
|
|
candidate: typing.Optional[CarPlateRegion] = None
|
|
for mask in _batchly_mask_car_plate(hsv):
|
|
# 连通域分析，筛选最符合车牌长宽比的区域作为车牌
|
|
candidate = _analyse_car_plate_connection(mask)
|
|
# 找到任意一个就退出
|
|
if candidate is not None:
|
|
break
|
|
|
|
if candidate is None:
|
|
logging.error("Can not find any car plate.")
|
|
return None
|
|
|
|
# 稍微扩边获取最终车牌区域
|
|
pad = 6
|
|
h_img, w_img = img.shape[:2]
|
|
x1 = max(candidate.x - pad, 0)
|
|
y1 = max(candidate.y - pad, 0)
|
|
x2 = min(candidate.x + candidate.w + pad, w_img)
|
|
y2 = min(candidate.y + candidate.h + pad, h_img)
|
|
logging.info(f"车牌区域: x={x1}, y={y1}, w={x2 - x1}, h={y2 - y1}")
|
|
|
|
# # 在原图上标记（仅供调试）
|
|
# debug = img.copy()
|
|
# cv.rectangle(debug, (x1, y1), (x2, y2), (0, 255, 0), 3)
|
|
# cv.imwrite('./debug_detected.jpg', debug)
|
|
|
|
# 二值化：文字/边缘 → 黑色，背景 → 白色
|
|
gray = cv.cvtColor(img[y1:y2, x1:x2], cv.COLOR_BGR2GRAY)
|
|
|
|
# 高斯模糊降噪
|
|
blurred = cv.GaussianBlur(gray, (3, 3), 0)
|
|
|
|
# Otsu 自动阈值，得到白字黑底，再取反 → 黑字白底
|
|
_, binary_otsu = cv.threshold(blurred, 0, 255, cv.THRESH_BINARY + cv.THRESH_OTSU)
|
|
# 反转：字符变黑，背景变白
|
|
binary = cv.bitwise_not(binary_otsu)
|
|
|
|
# 去除小噪点（开运算）
|
|
kernel_denoise = cv.getStructuringElement(cv.MORPH_RECT, (2, 2))
|
|
binary = cv.morphologyEx(binary, cv.MORPH_OPEN, kernel_denoise)
|
|
|
|
# 尝试获取视角矫正数据
|
|
perspective_data = _extract_perspective_data(gray)
|
|
if perspective_data is None:
|
|
logging.warning(f'Can not fetch perspective data. The output image has no perspective correction.')
|
|
return binary
|
|
|
|
# 执行视角矫正
|
|
perspective_src = np.array([
|
|
list(perspective_data.top_left),
|
|
list(perspective_data.top_right),
|
|
list(perspective_data.bottom_right),
|
|
list(perspective_data.bottom_left)
|
|
], dtype="float32")
|
|
perspective_dst = np.array([
|
|
[0, 0],
|
|
[perspective_data.new_width - 1, 0],
|
|
[perspective_data.new_width - 1, perspective_data.new_height - 1],
|
|
[0, perspective_data.new_height - 1]
|
|
], dtype="float32")
|
|
M = cv.getPerspectiveTransform(perspective_src, perspective_dst)
|
|
warped = cv.warpPerspective(binary, M, (perspective_data.new_width, perspective_data.new_height))
|
|
|
|
return warped
|
|
# cv.imwrite('./plate_binary.png', binary)
|
|
# print("二值化结果已保存: plate_binary.png")
|
|
|
|
# ── 4. 叠加边框轮廓（细化文字边缘，参考效果图）─────────────────────
|
|
# Canny 边缘叠加让效果更接近参考图
|
|
edges = cv.Canny(blurred, 40, 120)
|
|
edges_inv = cv.bitwise_not(edges) # 边缘→黑色
|
|
combined = cv.bitwise_and(binary, edges_inv) # 合并
|
|
# 再做一次轻微腐蚀让字体略粗
|
|
kernel_dilate = cv.getStructuringElement(cv.MORPH_RECT, (2, 2))
|
|
combined = cv.erode(combined, kernel_dilate, iterations=1)
|
|
|
|
return combined
|
|
# cv.imwrite('./plate_final.png', combined)
|
|
# print("最终结果已保存: plate_final.png")
|
|
|
|
|
|
@dataclass
|
|
class Cli:
|
|
input_file: Path
|
|
"""The path to input file"""
|
|
output_file: Path
|
|
"""The path to output file"""
|
|
|
|
@staticmethod
|
|
def from_cmdline() -> "Cli":
|
|
# Build parser
|
|
parser = argparse.ArgumentParser(
|
|
prog="Car Plate Extractor",
|
|
description="Extract the car plate part from given image.",
|
|
)
|
|
parser.add_argument(
|
|
"-i",
|
|
"--in",
|
|
required=True,
|
|
type=str,
|
|
action="store",
|
|
dest="input_file",
|
|
metavar="in.jpg",
|
|
help="""The path to input image containing car plate.""",
|
|
)
|
|
parser.add_argument(
|
|
"-o",
|
|
"--out",
|
|
required=True,
|
|
type=str,
|
|
action="store",
|
|
dest="output_file",
|
|
metavar="out.png",
|
|
help="""The path to output image for extracted car plate.""",
|
|
)
|
|
|
|
# Parse argument from cmdline and return
|
|
args = parser.parse_args()
|
|
return Cli(Path(args.input_file), Path(args.output_file))
|
|
|
|
|
|
def main():
|
|
# Setup logging format
|
|
logging.basicConfig(format="[%(levelname)s] %(message)s", level=logging.DEBUG)
|
|
|
|
# Get user request
|
|
cli = Cli.from_cmdline()
|
|
|
|
# Load file
|
|
in_img = cv.imread(str(cli.input_file), cv.IMREAD_COLOR)
|
|
if in_img is None:
|
|
logging.error(f"Fail to load image {cli.input_file}")
|
|
return
|
|
# Save extracted file if possible
|
|
out_img = extract_car_plate(in_img)
|
|
if out_img is not None:
|
|
cv.imwrite(str(cli.output_file), out_img)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
main()
|