基于openCV数字图像与机器视觉（转为HSV/HSI、将车牌数字分割为单个的字符图片）

一、彩色图像文件转为灰度文件

1. 使用opencv

代码：

import cv2 as cv
img = cv.imread('./pic/lena.png',1)
img_1 = cv.cvtColor(img,cv.COLOR_BGR2GRAY)
cv.imshow('gray',img_1)
cv.imshow('colour',img)
cv.waitKey(0)

效果：

2. 不使用opencv

代码：

from PIL import Image
I = Image.open('./pic/lena.png')
L = I.convert('L')
L.show()

效果：

二、将彩色图像转为HSV、HSI格式

1. 转HSV

HSV 格式： H 代表色彩，S 代表颜色的深浅，V 代表着颜色的明暗程度。

HSV 颜色空间可以很好地把颜色信息和亮度信息分开，将它们放在不同的通道中，减小了光线对于特定颜色识别的影响。
在阴影检测算法中经常需要将RGB格式的图像转化为HSV格式，对于阴影区域而言，它的色度和饱和度相对于原图像而言变化不大，主要是亮度信息变化较大，，将RGB格式转化为HSV格式，就可以得到H、S、V分量，从而得到色度、饱和度、亮度得值；
代码：

import cv2 as cvimg = cv.imread('./pic/lena.png', 1)
cv.imshow('original image', img)
hsv = cv.cvtColor(img, cv.COLOR_BGR2HSV)
cv.imshow('HSV format image', hsv)
cv.waitKey(0)

效果：

2. 转HSI

HSL (色相hue, 饱和度saturation, 亮度lightness/luminance),
也称HLS 或 HSI (I指intensity)
与 HSV非常相似，仅用亮度（lightness）替代了明度（brightness）。
人的视觉对亮度的敏感程度远强于对颜色浓淡的敏感程度，为了便于颜色处理和识别，人的市局系统经常采用HSI彩色空间，它比RGB空间更符合人的视觉特性。此外，由于HSI空间中亮度和色度具有可分离性，使得图像处理和机器视觉中大量灰度处理算法都可在HSI空间方便进行
HSI颜色空间：

代码：

import cv2
import numpy as npdef rgbtohsi(rgb_lwpImg):rows = int(rgb_lwpImg.shape[0])cols = int(rgb_lwpImg.shape[1])b, g, r = cv2.split(rgb_lwpImg)# 归一化到[0,1]b = b / 255.0g = g / 255.0r = r / 255.0hsi_lwpImg = rgb_lwpImg.copy()H, S, I = cv2.split(hsi_lwpImg)for i in range(rows):for j in range(cols):num = 0.5 * ((r[i, j]-g[i, j])+(r[i, j]-b[i, j]))den = np.sqrt((r[i, j]-g[i, j])**2+(r[i, j]-b[i, j])*(g[i, j]-b[i, j]))theta = float(np.arccos(num/den))if den == 0:H = 0elif b[i, j] <= g[i, j]:H = thetaelse:H = 2*3.14169265 - thetamin_RGB = min(min(b[i, j], g[i, j]), r[i, j])sum = b[i, j]+g[i, j]+r[i, j]if sum == 0:S = 0else:S = 1 - 3*min_RGB/sumH = H/(2*3.14159265)I = sum/3.0# 输出HSI图像，扩充到255以方便显示，一般H分量在[0,2pi]之间，S和I在[0,1]之间hsi_lwpImg[i, j, 0] = H*255hsi_lwpImg[i, j, 1] = S*255hsi_lwpImg[i, j, 2] = I*255return hsi_lwpImg
if __name__ == '__main__':rgb_lwpImg = cv2.imread("./pic/lena.png")hsi_lwpImg = rgbtohsi(rgb_lwpImg)cv2.imshow('lena.jpg', rgb_lwpImg)cv2.imshow('hsi_lwpImg', hsi_lwpImg)key = cv2.waitKey(0) & 0xFFif key == ord('q'):cv2.destroyAllWindows()

效果：

三、车牌数字分割为单个的字符图片

1.图片准备

2. 代码实现

1. 读取图片

file_path = "./pic/License/"
licenses = os.listdir(file_path)
for license in licenses:path = file_path+licenseoutput_path = "./pic/"+license # 图片输出路径# 如果该路径存在则删除if os.path.isdir(output_path):shutil.rmtree(output_path)# 创建文件夹os.mkdir(output_path)# 1.读取图片src = cv2.imread(path)img = src.copy()

2. 图片预处理

• 去除车牌螺丝点

 # 去除车牌上螺丝，将其替换为车牌底色cv2.circle(img, (145, 20), 10, (255, 0, 0), thickness=-1)cv2.circle(img, (430, 20), 10, (255, 0, 0), thickness=-1)cv2.circle(img, (145, 170), 10, (255, 0, 0), thickness=-1)cv2.circle(img, (430, 170), 10, (255, 0, 0), thickness=-1)cv2.circle(img, (180, 90), 10, (255, 0, 0), thickness=-1)

• 图片灰度处理

# 3.灰度gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)

• 高斯滤波

# 4.高斯滤波GSblurred = cv2.GaussianBlur(gray, (5, 5), 12) # 参数自行调节

• 二值化

# 5.将灰度图二值化设定阈值ret, thresh = cv2.threshold(GSblurred , 127, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)print("ret",ret)

• 闭运算

# 6. 闭运算kernel = np.ones((3, 3), int)closed = cv2.morphologyEx(thresh, cv2.MORPH_CLOSE, kernel,iterations=2)

• 分割字符

# 7.分割字符white = [] # 记录每一列的白色像素总和black = [] # ..........黑色.......height = thresh.shape[0]width = thresh.shape[1]white_max = 0black_max = 0# 计算每一列的黑白色像素总和for i in range(width):s = 0 # 这一列白色总数t = 0 # 这一列黑色总数for j in range(height):if thresh[j][i] == 255:s += 1if thresh[j][i] == 0:t += 1white_max = max(white_max, s)black_max = max(black_max, t)white.append(s)black.append(t)

• 分割图像

arg = False  # False表示白底黑字；True表示黑底白字if black_max > white_max:arg = True# 分割图像def find_end(start_):end_ = start_ + 1for m in range(start_ + 1, width - 1):if (black[m] if arg else white[m]) > (0.95 * black_max if arg else 0.95 * white_max):  # 0.95这个参数请多调整，对应下面的0.05end_ = mbreakreturn end_n = 1start = 1end = 2i=0;cj=[]while n < width - 2:n += 1if (white[n] if arg else black[n]) > (0.05 * white_max if arg else 0.05 * black_max):# 上面这些判断用来辨别是白底黑字还是黑底白字# 0.05这个参数请多调整，对应上面的0.95start = nend = find_end(start)n = endif end - start > 5:cj.append(thresh[1:height, start:end])cv2.imwrite(output_path + '/' + str(i) + '.jpg', cj[i])i = i + 1;

3. 输出结果

部分展示

4. 源码

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""项目主题：车牌检测将车牌数字分割为单个的字符图片
"""
import os
import shutilimport cv2
import numpy as npfile_path = "./pic/License/"
licenses = os.listdir(file_path)
for license in licenses:path = file_path+licenseoutput_path = "./pic/"+license# 如果该路径存在则删除if os.path.isdir(output_path):shutil.rmtree(output_path)# 创建文件夹os.mkdir(output_path)# 1.读取图片src = cv2.imread(path)img = src.copy()# 2.去除车牌上螺丝，将其替换为车牌底色cv2.circle(img, (145, 20), 10, (255, 0, 0), thickness=-1)cv2.circle(img, (430, 20), 10, (255, 0, 0), thickness=-1)cv2.circle(img, (145, 170), 10, (255, 0, 0), thickness=-1)cv2.circle(img, (430, 170), 10, (255, 0, 0), thickness=-1)cv2.circle(img, (180, 90), 10, (255, 0, 0), thickness=-1)# 3.灰度gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)# 4.高斯滤波GSblurred = cv2.GaussianBlur(gray, (5, 5), 12)# 5.将灰度图二值化设定阈值ret, thresh = cv2.threshold(GSblurred , 127, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)print("ret",ret)# 6. 闭运算kernel = np.ones((3, 3), int)closed = cv2.morphologyEx(thresh, cv2.MORPH_CLOSE, kernel,iterations=2)二值化ret, thresh = cv2.threshold(closed, 127, 255, cv2.THRESH_BINARY+ cv2.THRESH_OTSU)# 7.分割字符white = [] # 记录每一列的白色像素总和black = [] # ..........黑色.......height = thresh.shape[0]width = thresh.shape[1]white_max = 0black_max = 0# 计算每一列的黑白色像素总和for i in range(width):s = 0 # 这一列白色总数t = 0 # 这一列黑色总数for j in range(height):if thresh[j][i] == 255:s += 1if thresh[j][i] == 0:t += 1white_max = max(white_max, s)black_max = max(black_max, t)white.append(s)black.append(t)# print(s)# print(t)arg = False  # False表示白底黑字；True表示黑底白字if black_max > white_max:arg = True# 分割图像def find_end(start_):end_ = start_ + 1for m in range(start_ + 1, width - 1):if (black[m] if arg else white[m]) > (0.95 * black_max if arg else 0.95 * white_max):  # 0.95这个参数请多调整，对应下面的0.05end_ = mbreakreturn end_n = 1start = 1end = 2i=0;cj=[]while n < width - 2:n += 1if (white[n] if arg else black[n]) > (0.05 * white_max if arg else 0.05 * black_max):# 上面这些判断用来辨别是白底黑字还是黑底白字# 0.05这个参数请多调整，对应上面的0.95start = nend = find_end(start)n = endif end - start > 5:cj.append(thresh[1:height, start:end])cv2.imwrite(output_path + '/' + str(i) + '.jpg', cj[i])i += 1;

四、参考

https://www.cnblogs.com/datou-swag/articles/10672207.html
https://blog.csdn.net/qq_47281915/article/details/121705585