[人工智能]OpenCV入门【C++版】

1. OpenCV3.4.6安装包(含contrib)：https://pan.baidu.com/s/1KBD-fAO63p0s5ANYa5XcEQ 提取码：p7j0
2. resources资源：https://pan.baidu.com/s/1nkQ6iVV7IeeP4gTXvM_DyQ 提取码：ypvt

Chapter1 读取图片/视频/摄像头

从文件读取图片

• Mat cv::imread(const String &filename, int flags = IMREAD_COLOR)

从文件加载图像。函数imread从指定文件加载图像并返回。 如果无法读取图像（由于缺少文件、权限不正确、格式不受支持或无效），该函数将返回一个空矩阵（Mat::data==NULL ）。在彩色图像的情况下，解码图像的通道将以 B G R 顺序存储。

• void cv::imshow(cosnst String &winnanme, InputArray mat)

在指定窗口中显示图像。这个函数后面应该是 cv::waitKey 函数，它显示指定毫秒的图像。否则，它不会显示图像。例如，waitKey(0) 将无限显示窗口，直到有任何按键（适用于图像显示）。 waitKey(25) 将显示一帧 25 毫秒，之后显示将自动关闭。（如果你把它放在一个循环中读取视频，它会逐帧显示视频）

• int cv::waitKey(int delay = 0)

等待按下的键。函数 waitKey无限等待按键事件（当 delay≤0时）或延迟毫秒，当它为正时。由于操作系统在切换线程之间有最短时间，因此该函数不会完全等待延迟毫秒，它会至少等待延迟毫秒，具体取决于当时您计算机上正在运行的其他内容。如果在指定的时间过去之前没有按下任何键，则返回被按下键的代码或 -1。

#include <opencv2/imgcodecs.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/imgproc.hpp>
#include <iostream>

using namespace cv;
using namespace std;

int main()
{
	string path = "Resources/test.png";
	Mat img = imread(path);
	imshow("Image", img);
	waitKey(0); //显示图片不会一闪而过

	return 0;
}

从文件读取视频

要捕获视频，需要创建一个VideoCapture对象。它的参数可以是视频文件的名称或设备索引。
OpenCV3.4.6中VideoCapture类构造函数及成员函数

• cv::VideoCapture::VideoCapture()
• cv::VideoCapture::VideoCapture(const String &filename)
• cv::VideoCapture::VideoCapture(const String &filename, int apiPreference)
• cv::VideoCapture::VideoCapture(int index)
• cv::VideoCapture::VideoCapture(int index, int apiPreference)

打开视频文件或捕获设备或 IP 视频流进行视频捕获。

• virtual bool cv::VideoCapture::isOpened() const

如果视频捕获已经初始化，则返回true。如果先前对 VideoCapture构造函数或VideoCapture::open()的调用成功，则该方法返回 true。

• virtual bool cv::VideoCapture::read(OutputArray image)

抓取、解码并返回下一个视频帧。

• virtual double cv::VideoCapture::get(int proId) const

返回指定的VideoCapture属性。

• virtual double cv::VideoCapture::set(int proId, double value)

在VideoCapture中设置一个属性。

#include <opencv2/imgcodecs.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/imgproc.hpp>
#include <iostream>

using namespace cv;
using namespace std;

int main()
{
	string path = "Resources/test_video.mp4";
	VideoCapture cap(path); //视频捕捉对象
	Mat img;
	while (true) {

		cap.read(img);

		imshow("Image", img);
		waitKey(1);
	}
	return 0;
}

读摄像头

#include <opencv2/imgcodecs.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/imgproc.hpp>
#include <iostream>

using namespace cv;
using namespace std;

int main()
{
	VideoCapture cap(0);
	Mat img;

	while (true) {

		cap.read(img);

		imshow("Image", img);
		waitKey(1);
	}

	return 0;
}

Chapter2 基础函数

• void cv::cvtColor(InputArray src, OutputArray dst, int code, int dstCn = 0)

将图像从一种颜色空间转换为另一种颜色空间。该函数将输入图像从一种颜色空间转换为另一种颜色空间。在从 RGB 颜色空间转换的情况下，应明确指定通道的顺序（RGB 或 BGR）。man请注意，OpenCV 中的默认颜色格式通常称为 RGB，但实际上是 BGR（字节反转）。因此，标准（24 位）彩色图像中的第一个字节将是 8 位蓝色分量，第二个字节将是绿色，第三个字节将是红色。 然后第四、第五和第六个字节将是第二个像素（蓝色，然后是绿色，然后是红色），依此类推。

• void cv::GaussianBlur(InputArray src, OutputArray dst, Size ksize, double sigmaX, doube sigmaY = 0, int borderType = BORDER_DEFAULT)

使用高斯滤波器模糊图像。该函数将源图像与指定的高斯核进行卷积。

• void cv::Canny(InputArray image, OutputArray edges, double threshold1, double threshold2, int apertureSize = 3, bool L2gradient = false)

使用 Canny 算法在图像中查找边缘。

• Mat cv::getStructuringElement(int shape, Size ksize, Point anchor = Point(-1, -1))

返回指定大小和形状的结构元素，用于形态学操作。该函数构造并返回可以进一步传递给腐蚀、扩张或形态学的结构元素。 但是您也可以自己构建任意二进制掩码并将其用作结构元素。

• void cv::dilate(InputArray src, OutputArray dst, InuputArray kernel, Point anchor = Point(-1, -1), int iterations = 1, int borderType = BORDER_CONSTANT, const Scalar &borderValue = morphologyDefaultBorderValue())

使用特定的结构元素膨胀图像。

• void cv::erode(InputArray src, OutputArray dst, InuputArray kernel, Point anchor = Point(-1, -1), int iterations = 1, int borderType = BORDER_CONSTANT, const Scalar &borderValue = morphologyDefaultBorderValue())

使用特定的结构元素腐蚀图像。

#include <opencv2/imgcodecs.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/imgproc.hpp>
#include <iostream>

using namespace cv;
using namespace std;

int main()
{
	string path = "resources/test.png";
	Mat img = imread(path);
	Mat imgGray, imgBlur, imgCanny, imgDil, imgErode;

	cvtColor(img, imgGray, COLOR_BGR2GRAY); //灰度化
	GaussianBlur(img, imgBlur, Size(3, 3), 3, 0); //高斯模糊
	Canny(imgBlur, imgCanny, 25, 75); //边缘检测

	Mat kernel = getStructuringElement(MORPH_RECT, Size(3, 3));
	dilate(imgCanny, imgDil, kernel);
	erode(imgDil, imgErode, kernel);

	imshow("Image", img);
	imshow("ImageGray", imgGray);
	imshow("ImageBlur", imgBlur);
	imshow("ImageCanny", imgCanny);
	imshow("ImageDilation", imgDil);
	imshow("ImageErode", imgErode);
	waitKey(0);

	return 0;
}

Chapter3 调整和剪裁

• void cv::resize(InputArray src, OutputArray dst, Size dsize, double fx=0, double fy=0, int interpolation = INTER_LINEAR)

调整图像的大小。函数 resize 将图像 src 的大小缩小到或最大到指定的大小。请注意，不考虑初始 dst类型或大小。相反，大小和类型是从 src、dsize、fx 和 fy 派生的。

#include <opencv2/imgcodecs.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/imgproc.hpp>
#include <iostream>

using namespace cv;
using namespace std;

int main()
{
	string path = "resources/test.png";
	Mat img = imread(path);
	Mat imgResize, imgCrop;

	cout << img.size() << endl;
	resize(img, imgResize, Size(), 0.5, 0.5);

	Rect roi(200, 100, 300, 300);
	imgCrop = img(roi);

	imshow("Image", img);
	imshow("ImageResieze", imgResize);
	imshow("ImageCrop", imgCrop);
	waitKey(0);

	return 0;
}

Chapter4 绘制形状和文字

• Mat(int rows, int cols, int type, const Scalar &s)

重载的构造函数

• void cv::circle(InputOutputArray img, Point center, int radius, const Scalar &color, int thickness=1, int lineType=LINE_8, int shift=0)

函数 cv::circle 用给定的中心和半径绘制一个简单的或实心圆。

• void cv::rectangle(InputOutputArray img, Point pt1, Point pt2, const Scalar &color, int thickness=1, int lineType=LINE_8, int shift=0)
• void cv::rectangle(Mat &img, Rect rec, const Scalar &color, int thickness=1, int lineType=LINE_8, int shift=0)

绘制一个简单的、粗的或填充的右上矩形。函数 cv::rectangle 绘制一个矩形轮廓或两个对角为 pt1 和 pt2 的填充矩形。

• void cv::line (InputOutputArray img, Point pt1, Point pt2, const Scalar &color, int thickness=1, int lineType=LINE_8, int shift=0)

绘制连接两点的线段。函数line绘制图像中 pt1 和 pt2 点之间的线段。

• void cv::putText (InputOutputArray img, const String &text, Point org, int fontFace, double fontScale, Scalar color, int thickness=1, int lineType=LINE_8, bool bottomLeftOrigin=false)

绘制一个文本字符串。函数 cv::putText 在图像中呈现指定的文本字符串。无法使用指定字体呈现的符号将替换为问号。

#include <opencv2/imgcodecs.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/imgproc.hpp>
#include <iostream>

using namespace cv;
using namespace std;

int main()
{
	//Blank Image
	Mat img(512, 512, CV_8UC3, Scalar(255, 255, 255));

	circle(img, Point(256, 256), 155, Scalar(0, 69, 255), FILLED);
	rectangle(img, Point(130, 226), Point(382, 286), Scalar(255, 255, 255), -1);
	line(img, Point(130, 296), Point(382, 296), Scalar(255, 255, 255), 2);

	putText(img, "SJN's Workshop", Point(137, 262), FONT_HERSHEY_DUPLEX, 0.95, Scalar(0, 69, 255), 2);

	imshow("Image", img);
	waitKey(0);

	return 0;
}

Chapter5 透视变换

• Mat cv::getPerspectiveTransform (const Point2f src[], const Point2f dst[])

返回相应 4 个点对的 3x3 透视变换。

• void cv::warpPerspective (InputArray src, OutputArray dst, InputArray M, Size dsize, int flags=INTER_LINEAR, int borderMode=BORDER_CONSTANT, const Scalar &borderValue=Scalar())

对图像应用透视变换。

#include <opencv2/imgcodecs.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/imgproc.hpp>
#include <iostream>

using namespace cv;
using namespace std;

float w = 250, h = 350;
Mat matrix, imgWarp;

int main()
{
	string path = "Resources/cards.jpg";
	Mat img = imread(path);

	Point2f src[4] = { {529, 142}, {771, 190}, {405, 395}, {674, 457} };
	Point2f dst[4] = { {0.0f, 0.0f}, {w, 0.0f}, {0.0f, h}, {w, h} };

	matrix = getPerspectiveTransform(src, dst);
	warpPerspective(img, imgWarp, matrix, Point(w, h));

	for (int i = 0; i < 4; i++) {
		circle(img, src[i], 10, Scalar(0, 0, 255), FILLED);
	}

	imshow("Image", img);
	imshow("ImageWarp", imgWarp);
	waitKey(0);

	return 0;
}

注:文档扫描用到这种变换技术

Chapter6 颜色检测

• void cv::inRange (InputArray src, InputArray lowerb, InputArray upperb, OutputArray dst)

检查数组元素是否位于其他两个数组的元素之间。

• void cv::namedWindow (const String &winname, int flags = WINDOW_AUTOSIZE)

创建一个窗口。函数namedWindow创建一个可用作图像和轨迹栏占位符的窗口。创建的窗口由它们的名称引用。如果同名的窗口已经存在，则该函数不执行任何操作。

• int cv::createTrackbar (const String &trackbarname, const String &winname, int *value, int count, TrackbarCallback onChange = 0, void *userdata = 0)

创建一个trackbar并将其附加到指定窗口。函数createTrackbar创建一个具有指定名称和范围的trackbar（滑块或范围控件），分配一个变量值作为与trackbar同步的位置，并指定回调函数onChange为 在跟踪栏位置变化时被调用。创建的轨迹栏显示在指定的窗口winname中。

#include <opencv2/imgcodecs.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/imgproc.hpp>
#include <iostream>

using namespace cv;
using namespace std;

Mat imgHSV, mask;
int hmin = 0, smin = 110, vmin = 153;
int hmax = 19, smax = 240, vmax = 255;


int main()
{
	string path = "resources/lambo.png";
	Mat img = imread(path);
	cvtColor(img, imgHSV, COLOR_BGR2HSV);

	namedWindow("Trackbars", (640, 200));
	createTrackbar("Hue Min", "Trackbars", &hmin, 179);
	createTrackbar("Hue Max", "Trackbars", &hmax, 179);
	createTrackbar("Sat Min", "Trackbars", &smin, 255);
	createTrackbar("Sat Max", "Trackbars", &smax, 255);
	createTrackbar("Val Min", "Trackbars", &vmin, 255);
	createTrackbar("Val Max", "Trackbars", &vmax, 2555);


	while (true) {

		Scalar lower(hmin, smin, vmin);
		Scalar upper(hmax, smax, vmax);
		inRange(imgHSV, lower, upper, mask);

		imshow("Image", img);
		imshow("Image HSV", imgHSV);
		imshow("Image Mask", mask);
		waitKey(1);

	}

	return 0;
}

Chapter7 形状/轮廓检测

• void cv::findContours(InputOutputArray image, OutputArrayOfArrays contours, OutputArray hierarchy, int mode, int method, Point offset = Point())

在二值图像中查找轮廓。从OpenCV3.2开始源图像不会这个函数被修改。

• double cv::contourArea(InputArray contour, bool oriented=false)

计算轮廓区域

• double cv::arcLength(InputArray curve, bool closed)

计算曲线长度或闭合轮廓周长

• void cv::approxPolyDP(InputArray curve, OutputArray approxCurve, double epsilon, bool closed)

函数cv::approxPolyDP用另一个具有较少顶点的曲线/多边形来逼近一条曲线或多边形，以使它们之间的距离小于或等于指定的精度。

• Rect cv::boundingRect(InputArray array)

计算并返回指定点集或灰度图像非零像素的最小上边界矩形。

• void cv::drawContours(InputOutputArray image, InputArrayOfArrays contours, int contourIdx, const Scalar &color, int thickness = 1, int lineType = LINE_8, InputArray hierarchy = noArray(), int maxLevel = INT_MAX, Point offset = Point())

绘制轮廓轮廓或填充轮廓。如果厚度≥0，该函数在图像中绘制轮廓轮廓，如果厚度<0，则填充轮廓所包围的区域。

• Point_< _Tp > tl() const

左上角

• Point_< _Tp > br() const

右下角

//rect
template<typename _Tp> class cv::Rect_< _Tp >
typedef Rect_<int> cv::Rect2i
typedef Rect2i cv::Rect
//point
template<typename _Tp> class cv::Point_< _Tp >
typedef Point_<int> cv::Point2i
typedef Point2i cv::Point

#include <opencv2/imgcodecs.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/imgproc.hpp>
#include <iostream>

using namespace cv;
using namespace std;

void getContours(Mat imgDil, Mat img) {

	vector<vector<Point>> contours; //轮廓数据
	vector<Vec4i> hierarchy;

	findContours(imgDil, contours, hierarchy, RETR_EXTERNAL, CHAIN_APPROX_SIMPLE); //通过预处理的二值图像找到所有轮廓contours
	//drawContours(img, contours, -1, Scalar(255, 0, 255), 2); //绘制所有轮廓

	for (int i = 0; i < contours.size(); i++) 
	{
		double area = contourArea(contours[i]); //计算每个轮廓区域
		cout << area << endl;

		vector<vector<Point>> conPoly(contours.size()); 
		vector<Rect> boundRect(contours.size());
		string objectType;

		if (area > 1000) //过滤噪声
		{
			//找轮廓的近似多边形或曲线
			double peri = arcLength(contours[i], true);
			approxPolyDP(contours[i], conPoly[i], 0.02 * peri, true);
			
			cout << conPoly[i].size() << endl;
			boundRect[i] = boundingRect(conPoly[i]); //找每个近似曲线的最小上边界矩形
			
			int objCor = (int)conPoly[i].size();

			if (objCor == 3) { objectType = "Tri"; }
			if (objCor == 4) { 
				
				float aspRatio = (float)boundRect[i].width / boundRect[i].height; //宽高比
				cout << aspRatio << endl;
				if (aspRatio > 0.95 && aspRatio < 1.05) {
					objectType = "Square";
				}
				else {
					objectType = "Rect";
				}
			}
			if (objCor > 4) { objectType = "CirCle"; }

			drawContours(img, conPoly, i, Scalar(255, 0, 255), 2); //绘制滤除噪声后的所有轮廓
			rectangle(img, boundRect[i].tl(), boundRect[i].br(), Scalar(0, 255, 0), 5); //绘制边界框
			putText(img, objectType, { boundRect[i].x, boundRect[i].y - 5 }, FONT_HERSHEY_PLAIN, 1, Scalar(0, 69, 255), 1);
		}
	} 
}

int main()
{
	string path = "resources/shapes.png";
	Mat img = imread(path);
	Mat imgGray, imgBlur, imgCanny, imgDil;

	// Preprocessing
	cvtColor(img, imgGray, COLOR_BGR2GRAY);
	GaussianBlur(imgGray, imgBlur, Size(3, 3), 3, 0);
	Canny(imgBlur, imgCanny, 25, 75);
	Mat kernel = getStructuringElement(MORPH_RECT, Size(3, 3));
	dilate(imgCanny, imgDil, kernel);

	getContours(imgDil, img);

	imshow("Image", img);
	/*imshow("Image Gray", imgGray);
	imshow("Image Blur", imgBlur);
	imshow("Image Canny", imgCanny);
	imshow("Image Dil", imgDil);*/

	waitKey(0);

	return 0;
}

Chapter8 人脸检测

涉及模块objdetect:Object Detection

• class cv::CascadeClassifier

用于对象检测的级联分类器类。

• bool load (const String &filename)

从文件加载分类器。

• bool empty() const

检查分类器是否已加载。

• void detectMultiScale(InputArray image, std::vector<Rect> &objects, double scaleFactor=1.1, int minNeighbors=3, int flags=0, Size minSize=Size(), Size maxSize=Size())

检测输入图像中不同大小的对象。检测到的对象作为矩形列表返回。

#include <opencv2/imgcodecs.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/imgproc.hpp>
#include <opencv2/objdetect.hpp>
#include <iostream>

using namespace cv;
using namespace std;

int main()
{
	string path = "Resources/test.png";
	Mat img = imread(path);

	CascadeClassifier faceCascade;
	faceCascade.load("Resources/haarcascade_frontalface_default.xml");

	if (faceCascade.empty()) { cout << "XML file not loaded" << endl; }

	vector<Rect> faces;
	faceCascade.detectMultiScale(img, faces, 1.1, 10);

	for (int i = 0; i < faces.size(); i++) 
	{
		rectangle(img, faces[i].tl(), faces[i].br(), Scalar(255, 0, 255), 3);
	}

	imshow("Image", img);
	waitKey(0);

	return 0;
}

Project1 虚拟画家

颜色选择器：先找出待检测颜色的HSV阈值

#include <opencv2/imgcodecs.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/imgproc.hpp>
#include <iostream>
using namespace cv;
using namespace std;

int main()
{
	VideoCapture cap(1);
	Mat img;
	Mat imgHSV, mask, imgColor;
	int hmin = 0, smin = 0, vmin = 0;
	int hmax = 179, smax = 255, vmax = 255;

	namedWindow("Trackbars", (640, 200)); // Create Window
	createTrackbar("Hue Min", "Trackbars", &hmin, 179);
	createTrackbar("Hue Max", "Trackbars", &hmax, 179);
	createTrackbar("Sat Min", "Trackbars", &smin, 255);
	createTrackbar("Sat Max", "Trackbars", &smax, 255);
	createTrackbar("Val Min", "Trackbars", &vmin, 255);
	createTrackbar("Val Max", "Trackbars", &vmax, 255);

	while (true) {

		cap.read(img);
		cvtColor(img, imgHSV, COLOR_BGR2HSV);

		Scalar lower(hmin, smin, vmin);
		Scalar upper(hmax, smax, vmax);

		inRange(imgHSV, lower, upper, mask);
		// hmin, smin, vmin, hmax, smax, vmax;
		cout << hmin << ", " << smin << ", " << vmin << ", " << hmax << ", " << smax << ", " << vmax << endl;
		imshow("Image", img);
		imshow("Mask", mask);
		waitKey(1);
	}
}

利用检测到颜色的矩形框上边界中点开始虚拟作画

#include <opencv2/imgcodecs.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/imgproc.hpp>
#include <iostream>

using namespace cv;
using namespace std;

Mat img;
vector<vector<int>> newPoints;

vector<vector<int>> myColors{ {124, 48, 117, 143, 170, 255}, //purple
								{68, 72, 156, 102, 126, 255} }; //green

vector<Scalar> myColorValues{ {255, 0, 255}, //purple
								{0, 255, 0} }; //green

Point getContours(Mat imgDil) {

	vector<vector<Point>> contours; //轮廓数据
	vector<Vec4i> hierarchy;

	findContours(imgDil, contours, hierarchy, RETR_EXTERNAL, CHAIN_APPROX_SIMPLE); //通过预处理的二值图像找到所有轮廓contours
	//drawContours(img, contours, -1, Scalar(255, 0, 255), 2); //绘制所有轮廓(不滤除噪声)
	vector<vector<Point>> conPoly(contours.size());
	vector<Rect> boundRect(contours.size());
	Point myPoint(0, 0);

	for (int i = 0; i < contours.size(); i++)
	{
		double area = contourArea(contours[i]); //计算每个轮廓区域
		cout << area << endl;

		if (area > 1000) //过滤噪声
		{
			//找轮廓的近似多边形或曲线
			double peri = arcLength(contours[i], true);
			approxPolyDP(contours[i], conPoly[i], 0.02 * peri, true);

			cout << conPoly[i].size() << endl;
			boundRect[i] = boundingRect(conPoly[i]); //找每个近似曲线的最小上边界矩形
			myPoint.x = boundRect[i].x + boundRect[i].width / 2;
			myPoint.y = boundRect[i].y;

			//drawContours(img, conPoly, i, Scalar(255, 0, 255), 2); //绘制滤除噪声后的所有轮廓
			//rectangle(img, boundRect[i].tl(), boundRect[i].br(), Scalar(0, 255, 0), 5); //绘制边界框
		}
	}
	return myPoint; //返回矩形框上边界中点坐标
}

vector<vector<int>> findColor(Mat img)
{
	Mat imgHSV, mask;
	cvtColor(img, imgHSV, COLOR_BGR2HSV);

	for (int i = 0; i < myColors.size(); i++) 
	{
		Scalar lower(myColors[i][0], myColors[i][1], myColors[i][2]);
		Scalar upper(myColors[i][3], myColors[i][4], myColors[i][5]);
		inRange(imgHSV, lower, upper, mask);
		//imshow(to_string(i), mask);
		Point myPoint = getContours(mask); //根据mask得到检测到当前颜色矩形框的上边界中点坐标

		if (myPoint.x != 0 && myPoint.y != 0) 
		{
			newPoints.push_back({ myPoint.x, myPoint.y, i }); //得到当前帧检测颜色的目标点
		}
	}
	return newPoints;
}

void drawOnCanvas(vector<vector<int>> newPoints, vector<Scalar> myColorValues)
{
	for (int i = 0; i < newPoints.size(); i++) 
	{
		circle(img, Point(newPoints[i][0], newPoints[i][1]), 10, myColorValues[newPoints[i][2]], FILLED);
	}
}

int main()
{
	VideoCapture cap(0);

	while (true) 
	{
		cap.read(img);
		newPoints = findColor(img);
		drawOnCanvas(newPoints, myColorValues);

		imshow("Image", img);
		waitKey(1);
	}

	return 0;
}

Project2 文档扫描

#include <opencv2/imgcodecs.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/imgproc.hpp>
#include <iostream>

using namespace cv;
using namespace std;

Mat imgOriginal, imgGray, imgBlur,imgCanny, imgThre, imgDil, imgErode, imgWarp, imgCrop;
vector<Point> initialPoints, docPoints;

float w = 420, h = 596;


Mat preProcessing(Mat img)
{
	cvtColor(img, imgGray, COLOR_BGR2GRAY); 
	GaussianBlur(imgGray, imgBlur, Size(3, 3), 3, 0); 
	Canny(imgBlur, imgCanny, 25, 75); 

	Mat kernel = getStructuringElement(MORPH_RECT, Size(3, 3));
	dilate(imgCanny, imgDil, kernel);
	//erode(imgDil, imgErode, kernel);
	return imgDil;
}

vector<Point> getContours(Mat imgDil) {

	vector<vector<Point>> contours; //轮廓数据
	vector<Vec4i> hierarchy;

	findContours(imgDil, contours, hierarchy, RETR_EXTERNAL, CHAIN_APPROX_SIMPLE); //通过预处理的二值图像找到所有轮廓contours
	//drawContours(img, contours, -1, Scalar(255, 0, 255), 2); //绘制所有轮廓(不滤除噪声)
	vector<vector<Point>> conPoly(contours.size());
	vector<Point> biggest;
	int maxArea = 0;

	for (int i = 0; i < contours.size(); i++)
	{
		double area = contourArea(contours[i]); //计算每个轮廓区域
		cout << area << endl;

		if (area > 1000) //过滤噪声
		{
			//找轮廓的近似多边形或曲线
			double peri = arcLength(contours[i], true);
			approxPolyDP(contours[i], conPoly[i], 0.02 * peri, true);

			if (area > maxArea && conPoly[i].size() == 4) {

				//drawContours(imgOriginal, conPoly, i, Scalar(255, 0, 255), 5); //绘制滤除噪声后的所有轮廓
				biggest = { conPoly[i][0], conPoly[i][1], conPoly[i][2], conPoly[i][3] };
				maxArea = area;

			}
		}
	}
	return biggest; //返回最大轮廓四个点的坐标
}

void drawPoints(vector<Point> points, Scalar color)
{
	for (int i = 0; i < points.size(); i++) 
	{
		circle(imgOriginal, points[i], 10, color, FILLED);
		putText(imgOriginal, to_string(i), points[i], FONT_HERSHEY_PLAIN, 4, color, 4);
	}
}

vector<Point> reorder(vector<Point> points)
{
	vector<Point> newPoints;
	vector<int> sumPoints, subPoints;

	for (int i = 0; i < 4; i++) 
	{
		sumPoints.push_back(points[i].x + points[i].y);
		subPoints.push_back(points[i].x - points[i].y);
	}

	newPoints.push_back(points[min_element(sumPoints.begin(), sumPoints.end()) - sumPoints.begin()]); //0
	newPoints.push_back(points[max_element(subPoints.begin(), subPoints.end()) - subPoints.begin()]); //1
	newPoints.push_back(points[min_element(subPoints.begin(), subPoints.end()) - subPoints.begin()]); //2
	newPoints.push_back(points[max_element(sumPoints.begin(), sumPoints.end()) - sumPoints.begin()]); //3

	return newPoints;
}

Mat getWarp(Mat img, vector<Point> points, float w, float h)
{
	Point2f src[4] = { points[0], points[1], points[2], points[3] };
	Point2f dst[4] = { {0.0f, 0.0f}, {w, 0.0f}, {0.0f, h}, {w, h} };

	Mat matrix = getPerspectiveTransform(src, dst);
	warpPerspective(img, imgWarp, matrix, Point(w, h));
	return imgWarp;
}


int main()
{
	string path = "Resources/paper.jpg";
	imgOriginal = imread(path);
	//resize(imgOriginal, imgOriginal, Size(), 0.5, 0.5);

	//Preprocessing
	imgThre = preProcessing(imgOriginal);
	//Get Contours - Biggest
	initialPoints = getContours(imgThre);
	//drawPoints(initialPoints, Scalar(0, 0, 255));
	docPoints = reorder(initialPoints);
	//drawPoints(docPoints, Scalar(0, 255, 0));

	//Warp
	imgWarp = getWarp(imgOriginal, docPoints, w, h);

	//Crop
	int cropValue = 5;
	Rect roi(cropValue, cropValue, w - (2 * cropValue), h - (2 * cropValue));
	imgCrop = imgWarp(roi);

	imshow("Image", imgOriginal);
	imshow("Image Dilation", imgThre);
	imshow("Image Warp", imgWarp);
	imshow("Image Crop", imgCrop);
	waitKey(0);

	return 0;
}

Project3 车牌检测

#include <opencv2/imgcodecs.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/imgproc.hpp>
#include <opencv2/objdetect.hpp>
#include <iostream>

using namespace cv;
using namespace std;

int main()
{
	VideoCapture cap(0);
	Mat img;

	CascadeClassifier plateCascade;
	plateCascade.load("Resources/haarcascade_russian_plate_number.xml");

	if (plateCascade.empty()) { cout << "XML file not loaded" << endl; }

	vector<Rect> plates;

	while (true) {

		cap.read(img);

		plateCascade.detectMultiScale(img, plates, 1.1, 10);

		for (int i = 0; i < plates.size(); i++)
		{
			Mat imgCrop = img(plates[i]);
			imshow(to_string(i), imgCrop);
			imwrite("D:\\VS2019Projects\\chapter2\\chapter2\\resources\\Plates\\1.png", imgCrop);
			rectangle(img, plates[i].tl(), plates[i].br(), Scalar(255, 0, 255), 3);
		}

		imshow("Image", img);
		waitKey(1);
	}
	return 0;
}