opencv实现图像识别（数字图像处理库-OpenCV）
opencv实现图像识别（数字图像处理库-OpenCV）Imgproc核心模块，包含最基础的操作模块功能Core
1. OpenCV简介
OpenCV
OpenCV是一个基于BSD许可（开源）发行的跨平台计算机视觉和机器学习软件库，可以运行在Linux、Windows、Android和Mac OS操作系统上(未来期待在Harmony OS上运行）.
它轻量级而且高效——由一系列 C 函数和少量 C 类构成，同时提供了Python、Ruby、MATLAB等语言的接口，实现了图像处理和计算机视觉方面的很多通用算法。
2. Opencv模块
|
模块 |
功能 |
|
Core |
核心模块，包含最基础的操作 |
|
Imgproc |
图像处理模块 |
|
Objdectect |
目标检测模块 |
|
Feature2D |
2D特征检测模块 |
|
Video |
视频处理模块 |
|
HighGUI |
高层图像用户界面 |
|
Calib3d |
3D重建模块 |
|
ML |
机器学习模块 |
|
FLANN |
最近邻搜索模块 |
|
Stitching |
图像拼接模块 |
|
Photo |
计算图像学 |
|
Superres |
超分辨率模块 |
|
GPU |
GPU并行加速模块 |
3. OpenCV总览
OpenCV框架中的每一个模块都包含大量的计算机视觉方法，每一个模块都能独当一面，功能强大。
本篇文章将介绍OpenCV库中最重要的模块：Imgproc（图像处理模块）。
图像处理模块包括：图像的读取、显示、保存；几何运算；灰度变换；几何变换；平滑、锐化；数学形态学；阈值分割；边缘检测；色彩空间；形状绘制等。
图像读取、显示、保存
几何运算
灰度变换
几何变换
平滑、锐化
数学形态学
阈值分割
边缘检测
色彩空间
形状绘制
- 图像读取、显示、保存
|
函数 |
功能 |
|
cv2.imread( ) |
图像读取 |
|
cv2.imshow( ) |
图像显示 |
|
cv2.imwrite( ) |
图像保存 |
"""图像读取、显示、保存"""
img = cv2.imread('shiyuan.png')
cv2.imwrite('shi.png' img)
cv2.imshow('img' img)
cv2.waitKey(0)
cv2.destroyAllWindows()
- 几何运算
|
函数 |
功能 |
|
img1 img2 |
图像加法 |
|
cv2.addWeight( ) |
图像融合 |
"""几何运算"""
img1 = cv2.imread('shiyuan.png')
img2 = cv2.imread('lizi.png')
img3 = cv2.resize(img1 (300 300)) cv2.resize(img2 (300 300))
img4 = cv2.addWeighted(cv2.resize(img1 (300 300)) 0.3 cv2.resize(img2 (300 300)) 0.7 20)
cv2.imshow('img3' img3)
cv2.imshow('img4' img4)
cv2.waitKey(0)
cv2.destroyAllWindows()
- 灰度变换
|
函数 |
功能 |
|
对数变换 |
变换图像灰度 |
|
伽马变换 |
变换图像灰度 |
|
直方图均衡化 |
变换图像灰度 |
|
直方图规定化 |
变换图像灰度 |
"""灰度变换"""
import cv2
import copy
img = cv2.imread('bai.png' 1)
img1 = cv2.imread('bai.png' 0)
img = cv2.resize(img (400 300))
img1 = cv2.resize(img (400 300))
gray=cv2.cvtColor(img cv2.COLOR_BGR2GRAY)
#伽马变换
gamma = copy.deepcopy(gray)
rows = img.shape[0]
cols = img.shape[1]
for i in range(rows):
for j in range(cols):
gamma[i][j]=3*pow(gamma[i][j] 0.8)
cv2.imshow('img' img)
cv2.imshow('gray' img1)
cv2.imshow('gamma' gamma)
cv2.waitKey(0)
cv2.destroyAllWindows()
"""灰度变换"""
import cv2
import copy
import math
img = cv2.imread('bai.png' 1)
img1 = cv2.imread('bai.png' 0)
img = cv2.resize(img (400 300))
img1 = cv2.resize(img (400 300))
gray=cv2.cvtColor(img cv2.COLOR_BGR2GRAY)
#对数变换
logc = copy.deepcopy(gray)
rows=img.shape[0]
cols=img.shape[1]
for i in range(rows):
for j in range(cols):
logc[i][j] = 3 * math.log(1 logc[i][j])
cv2.imshow('img' img)
cv2.imshow('gray' img1)
cv2.imshow('logc' logc)
cv2.waitKey(0)
cv2.destroyAllWindows()
"""灰度变换"""
import cv2
import copy
import math
img = cv2.imread('bai.png' 1)
img1 = cv2.imread('bai.png' 0)
img = cv2.resize(img (400 300))
img1 = cv2.resize(img (400 300))
gray=cv2.cvtColor(img cv2.COLOR_BGR2GRAY)
# 反色变换
cover=copy.deepcopy(gray)
rows=img.shape[0]
cols=img.shape[1]
for i in range(rows):
for j in range(cols):
cover[i][j]=255-cover[i][j]
#通过窗口展示图片 第一个参数为窗口名 第二个为读取的图片变量
cv2.imshow('img' img)
cv2.imshow('gray' img1)
cv2.imshow('cover' cover)
cv2.waitKey(0)
cv2.destroyAllWindows()
#直方图规定化
import cv2
import numpy as np
import matplotlib.pyplot as plt
img0=cv2.imread('hua.png')#读取原图片
scr=cv2.imread('tu.png')#读取目标图片
#把两张图片转成真正的灰度图片，因为自己只会做灰度图片的规定化
img0=cv2.cvtColor(img0 cv2.COLOR_BGR2GRAY)
img=img0.copy()#用于之后做对比图
scr=cv2.cvtColor(scr cv2.COLOR_BGR2GRAY)
mHist1=[]
mNum1=[]
inhist1=[]
mHist2=[]
mNum2=[]
inhist2=[]
#对原图像进行均衡化
for i in range(256):
mHist1.append(0)
row col=img.shape#获取原图像像素点的宽度和高度
for i in range(row):
for j in range(col):
mHist1[img[i j]]= mHist1[img[i j]] 1#统计灰度值的个数
mNum1.append(mHist1[0]/img.size)
for i in range(0 255):
mNum1.append(mNum1[i] mHist1[i 1]/img.size)
for i in range(256):
inhist1.append(round(255*mNum1[i]))
#对目标图像进行均衡化
for i in range(256):
mHist2.append(0)
rows cols=scr.shape#获取目标图像像素点的宽度和高度
for i in range(rows):
for j in range(cols):
mHist2[scr[i j]]= mHist2[scr[i j]] 1#统计灰度值的个数
mNum2.append(mHist2[0]/scr.size)
for i in range(0 255):
mNum2.append(mNum2[i] mHist2[i 1]/scr.size)
for i in range(256):
inhist2.append(round(255*mNum2[i]))
- 几何变换
|
函数 |
功能 |
|
cv2.resize( ) |
图像缩放 |
|
cv2.warpAffine( ) |
图像平移 |
|
cv2.getRotationMatrix2D( ) cv2.warpAffine( ) |
图像旋转 |
|
cv2.getAffineTransform( ) |
仿射变换 |
|
cv2.getPerspectiveTransform( ) cv2.warpPerspective( ) |
透射变换 |
|
cv2.pyrUp( ) |
高斯金字塔上采样 |
|
cv2.pyrDown( ) |
高斯金字塔下采样 |
|
img-cv2.pyrUp(cv2.pyrDown(img)) |
拉普拉斯金字塔 |
"""几何变换"""
img = cv2.imread('shiyuan.png')
img1 = cv2.resize(img (300 300))
M = np.float32([[1 0 30] [0 1 60]])
img2 = cv2.warpAffine(img1 M (300 300))
img2 = cv2.putText(img2 'panning' (20 30) cv2.FONT_HERSHEY_SIMPLEX 1 (0 255 0) 2)
M = cv2.getRotationMatrix2D(((300-1)/2.0 (300-1)/2.0) 45 1)
img3 = cv2.warpAffine(img1 M (300 300))
img3 = cv2.putText(img3 'rotation' (20 30) cv2.FONT_HERSHEY_SIMPLEX 1 (0 255 0) 2)
matr1 = np.float32([[50 50] [200 50] [50 200]])
matr2 = np.float32([[10 100] [200 50] [100 250]])
M = cv2.getAffineTransform(matr1 matr2)
img4 = cv2.warpAffine(img1 M (300 300))
img4 = cv2.putText(img4 'affine' (20 30) cv2.FONT_HERSHEY_SIMPLEX 1 (0 255 0) 2)
matr1 = np.float32([[56 65] [368 52] [28 387] [389 390]])
matr2 = np.float32([[0 0] [300 0] [0 300] [300 300]])
M = cv2.getPerspectiveTransform(matr1 matr2)
img5 = cv2.warpPerspective(img1 M (300 300))
img5 = cv2.putText(img5 'perspective' (20 30) cv2.FONT_HERSHEY_SIMPLEX 1 (0 255 0) 2)
cv2.imshow('img1' img1)
cv2.imshow('img2' img2)
cv2.imshow('img3' img3)
cv2.imshow('img4' img4)
cv2.imshow('img5' img5)
cv2.waitKey(0)
cv2.destroyAllWindows()
"""图像金字塔"""
import cv2
#高斯金字塔
def pyramid_demo(image):
level = 2
temp = image.copy()
pyramid_images = []
for i in range(level):
dst = cv2.pyrDown(temp)
pyramid_images.append(dst)
cv2.imshow("pyramid" str(i 1) dst)
temp = dst.copy()
return pyramid_images
#拉普拉斯金字塔
def lapalian_demo(image):
pyramid_images = pyramid_demo(image)
level = len(pyramid_images)
for i in range(level-1 -1 -1):
if (i-1) < 0:
expand = cv2.pyrUp(pyramid_images[i] dstsize = image.shape[:2])
lpls = cv2.subtract(image expand)
cv2.imshow("lapalian_down_" str(i 1) lpls)
else:
expand = cv2.pyrUp(pyramid_images[i] dstsize = pyramid_images[i-1].shape[:2])
lpls = cv2.subtract(pyramid_images[i-1] expand)
cv2.imshow("lapalian_down_" str(i 1) lpls)
src = cv2.resize(cv2.imread('shiyuan.png') (256 256))
cv2.namedWindow('input_image')
cv2.imshow('input_image' src)
lapalian_demo(src)
cv2.waitKey(0)
cv2.destroyAllWindows()
"""直方图均衡化"""
import cv2
import numpy as np
img = cv2.imread('bai.png' 0)
img = cv2.resize(img (400 300))
equ = cv2.equalizeHist(img)
cv2.imshow('img' equ)
cv2.waitKey()
cv2.destroyAllWindows()
- 平滑、锐化
|
函数 |
功能 |
|
cv2.blur( ) |
均值滤波 |
|
cv2.GaussianBlur( ) |
高斯滤波 |
|
cv2.medianBlur( ) |
中值滤波 |
|
cv2.bilateralFilter( ) |
双边滤波 |
"""平滑、锐化"""
import cv2
img = cv2.imread('shiyuan.png')
img = cv2.resize(img (300 300))
img1 = cv2.blur(img (11 11))
img2 = cv2.GaussianBlur(img (11 11) 0)
img3 = cv2.medianBlur(img 11)
img4 = cv2.bilateralFilter(img 9 75 75)
M = np.ones((5 5) np.float32) / 25
img5 = cv.filter2D(img -1 M)
cv2.imshow('img1' img1)
cv2.imshow('img2' img2)
cv2.imshow('img3' img3)
cv2.imshow('img4' img4)
cv2.imshow('img5' img5)
cv2.waitKey(0)
cv2.destroyAllWindows()
- 数学形态学
|
函数 |
功能 |
|
cv2.erode( ) |
腐蚀 |
|
开运算 | |
|
cv2.morphologyEx( cv2.MORPH_OPEN) |
闭运算 |
|
cv2.morphologyEx( cv2.MORPH_CLOSE) |
顶帽运算 |
|
cv2.morphologyEx( cv2.MORPH_TOPHAT) |
底帽运算 |
|
cv2.morphologyEx( cv2.MORPH_BLACKHAT) |
形态学梯度 |
|
cv2.dilate( ) |
膨胀 |
"数学形态学"
import cv2
img = cv2.imread('shiyuan.png')
img = cv2.resize(img (300 300))
kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE (3 3))
img1 = cv2.dilate(img kernel)
img2 = cv2.erode(img kernel)
#设置结构元
kernel_rect=cv2.getStructuringElement(cv2.MORPH_RECT (3 3))
kernel_cross=cv2.getStructuringElement(cv2.MORPH_CROSS (3 3))
kernel_ellipse=cv2.getStructuringElement(cv2.MORPH_ELLIPSE (3 3))
#图像开运算处理
open_rect=cv2.morphologyEx(img cv2.MORPH_OPEN kernel_rect)
open_cross=cv2.morphologyEx(img cv2.MORPH_OPEN kernel_cross)
open_ellipse=cv2.morphologyEx(img cv2.MORPH_OPEN kernel_ellipse)
#图像闭运算处理
close_rect=cv2.morphologyEx(img cv2.MORPH_CLOSE kernel_rect)
close_cross=cv2.morphologyEx(img cv2.MORPH_CLOSE kernel_cross)
close_ellipse=cv2.morphologyEx(img cv2.MORPH_CLOSE kernel_ellipse)
gradient_rect = cv2.morphologyEx(img cv2.MORPH_GRADIENT kernel_rect)
gradient_cross = cv2.morphologyEx(img cv2.MORPH_GRADIENT kernel_cross)
gradient_ellipse = cv2.morphologyEx(img cv2.MORPH_GRADIENT kernel_ellipse)
#顶帽变换
tophat_rect=cv2.morphologyEx(img cv2.MORPH_TOPHAT kernel_rect)
tophat_cross=cv2.morphologyEx(img cv2.MORPH_TOPHAT kernel_cross)
tophat_ellipse=cv2.morphologyEx(img cv2.MORPH_TOPHAT kernel_ellipse)
#顶帽变换
blackhat_rect=cv2.morphologyEx(img cv2.MORPH_BLACKHAT kernel_rect)
blackhat_cross=cv2.morphologyEx(img cv2.MORPH_BLACKHAT kernel_cross)
blackhat_ellipse=cv2.morphologyEx(img cv2.MORPH_BLACKHAT kernel_ellipse)
cv2.imshow('blackhat_rect' blackhat_rect)
cv2.imshow('blackhat_cross' blackhat_cross)
cv2.imshow('blackhat_ellipse' blackhat_ellipse)
cv2.imshow('tophat_rect' tophat_rect)
cv2.imshow('tophat_cross' tophat_cross)
cv2.imshow('tophat_ellipse' tophat_ellipse)
cv2.imshow('img1' img1)
cv2.imshow('img2' img2)
cv2.imshow('open_rect' open_rect)
cv2.imshow('open_cross' open_cross)
cv2.imshow('open_ellipse' open_ellipse)
cv2.imshow('close_rect' close_rect)
cv2.imshow('close_cross' close_cross)
cv2.imshow('close_ellipse' close_ellipse)
cv2.imshow('gradient_rect' gradient_rect)
cv2.imshow('gradient_cross' gradient_cross)
cv2.imshow('gradient_ellipse' gradient_ellipse)
cv2.waitKey(0)
cv2.destroyAllWindows()
- 阈值分割
|
函数 |
功能 |
|
cv2.threshold( cv2.THRESH_BINARY) |
二值化阈值 |
|
cv2.threshold( cv2.THRESH_BINARY_INV) |
反二值化阈值 |
|
cv2.threshold( cv2.THRESH_TOZERO) |
低阈值零处理 |
|
cv2.threshold( cv2.THRESH_TOZERO_INV) |
超阈值零处理 |
|
cv2.threshold( cv2.THRESH_OSTU) |
大津算法 |
|
cv2.threshold( cv2.THRESH_TRIANGLE) |
截断阈值化处理 |
|
cv2.adaptiveThreshold( cv2.ADAPTIVE_THRESH_GAUSSIAN_C ) |
自适应阈值处理 |
|
cv2.adaptiveThreshold( cv2.ADAPTIVE_THRESH_MEAN_C ) |
自适应阈值处理 |
"阈值分割"
import cv2
img = cv2.imread('shiyuan.png')
img = cv2.resize(img (400 300))
img = cv2.cvtColor(img cv2.COLOR_BGR2GRAY)
ret img1 = cv2.threshold(img 110 255 cv2.THRESH_BINARY)
ret img2 = cv2.threshold(img 110 255 cv2.THRESH_BINARY_INV)
ret img3 = cv2.threshold(img 110 255 cv2.THRESH_TOZERO)
ret img4 = cv2.threshold(img 110 255 cv2.THRESH_TOZERO_INV)
ret img5 = cv2.threshold(img 110 255 cv2.THRESH_TRUNC)
ret img6 = cv2.threshold(img 110 255 cv2.THRESH_TRIANGLE)
ret img7 = cv2.threshold(img 110 255 cv2.THRESH_OTSU)
ret img8 = cv2.threshold(cv2.GaussianBlur(img (7 7) 0) 110 255 cv2.THRESH_OTSU)
img9 = cv2.adaptiveThreshold(img 127 cv2.ADAPTIVE_THRESH_GAUSSIAN_C cv2.THRESH_BINARY 9 11)
img10 = cv2.adaptiveThreshold(img 127 cv2.ADAPTIVE_THRESH_MEAN_C cv2.THRESH_BINARY 9 11)
cv2.imshow('img' img)
cv2.imshow('img1' img1)
cv2.imshow('img2' img2)
cv2.imshow('img3' img3)
cv2.imshow('img4' img4)
cv2.imshow('img5' img5)
cv2.imshow('img6' img6)
cv2.imshow('img7' img7)
cv2.imshow('img8' img8)
cv2.imshow('img9' img9)
cv2.imshow('img10' img10)
cv2.waitKey(0)
cv2.destroyAllWindows()
- 边缘检测
|
函数 |
功能 |
|
cv2.Canny( ) |
Canny算子 |
|
cv2.findContours( ) |
轮廓检测 |
|
cv2.filter2D( ) |
边缘提取 |
"边缘检测"
import cv2
img = cv2.imread('bai.png')
img = cv2.resize(img (400 300))
img1 = cv2.Canny(img 123 5)
cv2.imshow('img1' img1)
cv2.waitKey(0)
cv2.destroyAllWindows()
"""边缘检测"""
import cv2
img = cv2.imread('bai.png')
img = cv2.resize(img (400 300))
gray = cv2.cvtColor(img cv2.COLOR_BGR2GRAY)
ret binary = cv2.threshold(gray 127 255 cv2.THRESH_BINARY)
contours hierarchy = cv2.findContours(binary cv2.RETR_TREE cv2.CHAIN_APPROX_SIMPLE)
cv2.drawContours(img contours -1 (0 0 255) 1)
cv2.imshow("img" img)
cv2.waitKey(0)
cv2.destroyAllWindows()
"""边缘检测"""
import cv2
import numpy as np
def find_contours(kernel):
img = cv2.imread('bai.png')
img = cv2.resize(img (400 300))
img1 = cv2.filter2D(img -1 kernel)
cv2.imshow('img1' img1)
cv2.waitKey(0)
cv2.destroyAllWindows()
kernel1 = np.array((
[0.0625 0.125 0.0625]
[0.125 0.25 0.125]
[0.0625 0.125 0.0625]) dtype="float32")
#Sobel算子
kernel2 = np.array(([-1 -2 -1]
[0 0 0]
[1 2 1]))
kernel3 = np.array(([-2 -1 0]
[-1 1 1]
[0 -1 -2]))
kernel4 = np.array([[-1 -1 -1]
[-1 8 -1]
[-1 -1 -1]])
kernel5 = np.array([[0 -1 0]
[-1 5 -1]
[0 -1 0]])
kernel6 = np.array([[0 1 0]
[1 -4 1]
[0 1 0]])
find_contours(kernel1)
find_contours(kernel2)
find_contours(kernel3)
find_contours(kernel4)
find_contours(kernel5)
find_contours(kernel6)
- 色彩空间
|
函数 |
功能 |
|
cv2.cvtColor( cv2.COLOR_BGR2GRAY) |
图像灰度化 |
|
cv2.cvtColor( cv2.COLOR_BGR2HSV) |
RGB转HSV |
"""色彩空间"""
import cv2
img = cv2.imread('bai.png')
img = cv2.resize(img (400 300))
img1 = cv2.cvtColor(img cv2.COLOR_BGR2GRAY)
img2 = cv2.cvtColor(img cv2.COLOR_BGR2HSV)
cv2.imshow('img1' img1)
cv2.imshow('img2' img2)
cv2.waitKey(0)
cv2.destroyAllWindows()
- 形状绘制
|
函数 |
功能 |
|
cv2.line( ) |
绘制直线 |
|
cv2.circle( ) |
绘制圆圈 |
|
cv2.ellipse( ) |
绘制椭圆 |
|
cv2.rectangle( ) |
绘制矩形 |
|
cv2.arrowedLine( ) |
绘制箭头 |
|
cv2.putText( ) |
绘制文本 |
"""形状绘制"""
import cv2
img = cv2.imread('bai.png')
img = cv2.resize(img (400 300))
imgx = img.copy()
imgy = img.copy()
imgz = img.copy()
imgw = img.copy()
img = cv2.resize(img (400 300))
img1 = cv2.line(img (10 10) (200 300) (0 0 255) 2)
img2 = cv2.circle(imgx (60 60) 30 (0 0 213) -1)
img3 = cv2.rectangle(imgy (10 10) (100 80) (0 0 200) 2)
img4 = cv2.ellipse(imgz (256 256) (50 40) 0 5 360 (20 213 79) -1)
font=cv2.FONT_HERSHEY_SIMPLEX
img5 = cv2.putText(imgw 'opencv' (80 90) font 2 (255 255 255) 3)
cv2.imshow('img1' img1)
cv2.imshow('img2' img2)
cv2.imshow('img3' img3)
cv2.imshow('img4' img4)
cv2.imshow('img5' img5)
cv2.waitKey(0)
cv2.destroyAllWindows()
写在最后
资料包
下一期将扒拉sklearn库，sklearn库是做机器学习的不二之选，欢迎大家搬好小板凳呀！
