const int classNum = 10;  //總共有0~9個(gè)數(shù)字類別
	const int picNum = 20;//每個(gè)類別共20張圖片
	const int pic_w = 28;//圖片寬
	const int pic_h = 28;//圖片高

	//將數(shù)據(jù)集分為訓(xùn)練集、測(cè)試集
	double totalNum = classNum * picNum;//圖片總數(shù)
	double per = 0.8;	//百分比--修改百分比可改變訓(xùn)練集、測(cè)試集比重
	double trainNum = totalNum * per;//訓(xùn)練圖片數(shù)量
	double testNum = totalNum * (1.0 - per);//測(cè)試圖片數(shù)量

	Mat Train_Data, Train_Label;//用于訓(xùn)練
	vector<MyNum>TestData;//用于測(cè)試
	for (int i = 0; i < picNum; i++)
	{
		for (int j = 0; j < classNum; j++)
		{
			//將所有圖片數(shù)據(jù)都拷貝到Mat矩陣?yán)?
			Mat temp;
			gray(Range(j*pic_w, j*pic_w + pic_w), Range(i*pic_h, i*pic_h + pic_h)).copyTo(temp);
			Train_Data.push_back(temp.reshape(0, 1)); //將temp數(shù)字圖像reshape成一行數(shù)據(jù)，然后一一追加到Train_Data矩陣中
			Train_Label.push_back(j);

			//而外用于測(cè)試
			if (i * classNum + j >= trainNum)
			{
				TestData.push_back({ temp,Rect(i*pic_w,j*pic_h,pic_w,pic_h),j });
			}
		}
	}

    //準(zhǔn)備訓(xùn)練數(shù)據(jù)集
    Train_Data.convertTo(Train_Data, CV_32FC1); //轉(zhuǎn)化為CV_32FC1類型
    Train_Label.convertTo(Train_Label, CV_32FC1);
    Mat TrainDataMat = Train_Data(Range(0, trainNum), Range::all()); //只取trainNum行訓(xùn)練
    Mat TrainLabelMat = Train_Label(Range(0, trainNum), Range::all());

    //KNN訓(xùn)練
    const int k = 3;  //k值，取奇數(shù)，影響最終識(shí)別率
    Ptr<KNearest>knn = KNearest::create();  //構(gòu)造KNN模型
    knn->setDefaultK(k);//設(shè)定k值
    knn->setIsClassifier(true);//KNN算法可用于分類、回歸。
    knn->setAlgorithmType(KNearest::BRUTE_FORCE);//字符匹配算法
    knn->train(TrainDataMat, ROW_SAMPLE, TrainLabelMat);//模型訓(xùn)練

	//預(yù)測(cè)及結(jié)果顯示
	double count = 0.0;
	Scalar color;
	for (int i = 0; i < TestData.size(); i++)
	{
		//將測(cè)試圖片轉(zhuǎn)成CV_32FC1，單行形式
		Mat data = TestData[i].mat.reshape(0, 1);
		data.convertTo(data, CV_32FC1);
		Mat sample = data(Range(0, data.rows), Range::all());

		float f = knn->predict(sample); //預(yù)測(cè)
		if (f == TestData[i].label)
		{
			color = Scalar(0, 255, 0); //如果預(yù)測(cè)正確，繪制綠色，并且結(jié)果+1
			count++;
		}
		else
		{
			color = Scalar(0, 0, 255);//如果預(yù)測(cè)錯(cuò)誤，繪制紅色
		}

		rectangle(src, TestData[i].rect, color, 2);
	}

	//將繪制結(jié)果拷貝到一張新圖上
	Mat result(Size(src.cols, src.rows + 50), CV_8UC3, Scalar::all(255));
	src.copyTo(result(Rect(0, 0, src.cols, src.rows)));
	//將得分在結(jié)果圖上顯示
	char text[10];
	int score = (count / testNum) * 100;
	sprintf_s(text, "%s%d%s", "Score:", score, "%");
	putText(result, text, Point((result.cols / 2) - 80, result.rows - 15), FONT_HERSHEY_SIMPLEX, 1, Scalar(0, 255, 0), 2);
	

#include<iostream>
#include<opencv2/opencv.hpp>
#include<opencv2/ml.hpp>
using namespace std;
using namespace cv;
using namespace cv::ml;


//**自定義結(jié)構(gòu)體
struct MyNum
{
	cv::Mat mat; //數(shù)字圖片
	cv::Rect rect;//相對(duì)整張圖所在矩形
	int label;//數(shù)字標(biāo)簽
};

int main()
{
	Mat src = imread("digit.png");
	if (src.empty())
	{
		cout << "No Image..." << endl;
		system("pause");
		return -1;
	}

	Mat gray;
	cvtColor(src, gray, COLOR_BGR2GRAY);

	const int classNum = 10;  //總共有0~9個(gè)數(shù)字類別
	const int picNum = 20;//每個(gè)類別共20張圖片
	const int pic_w = 28;//圖片寬
	const int pic_h = 28;//圖片高

	//將數(shù)據(jù)集分為訓(xùn)練集、測(cè)試集
	double totalNum = classNum * picNum;//圖片總數(shù)
	double per = 0.8;	//百分比--修改百分比可改變訓(xùn)練集、測(cè)試集比重
	double trainNum = totalNum * per;//訓(xùn)練圖片數(shù)量
	double testNum = totalNum * (1.0 - per);//測(cè)試圖片數(shù)量

	Mat Train_Data, Train_Label;//用于訓(xùn)練
	vector<MyNum>TestData;//用于測(cè)試
	for (int i = 0; i < picNum; i++)
	{
		for (int j = 0; j < classNum; j++)
		{
			//將所有圖片數(shù)據(jù)都拷貝到Mat矩陣?yán)?
			Mat temp;
			gray(Range(j*pic_w, j*pic_w + pic_w), Range(i*pic_h, i*pic_h + pic_h)).copyTo(temp);
			Train_Data.push_back(temp.reshape(0, 1)); //將temp數(shù)字圖像reshape成一行數(shù)據(jù)，然后一一追加到Train_Data矩陣中
			Train_Label.push_back(j);

			//額外用于測(cè)試
			if (i * classNum + j >= trainNum)
			{
				TestData.push_back({ temp,Rect(i*pic_w,j*pic_h,pic_w,pic_h),j });
			}
		}
	}

	//準(zhǔn)備訓(xùn)練數(shù)據(jù)集
	Train_Data.convertTo(Train_Data, CV_32FC1); //轉(zhuǎn)化為CV_32FC1類型
	Train_Label.convertTo(Train_Label, CV_32FC1);
	Mat TrainDataMat = Train_Data(Range(0, trainNum), Range::all()); //只取trainNum行訓(xùn)練
	Mat TrainLabelMat = Train_Label(Range(0, trainNum), Range::all());

	//KNN訓(xùn)練
	const int k = 3;  //k值，取奇數(shù)，影響最終識(shí)別率
	Ptr<KNearest>knn = KNearest::create();  //構(gòu)造KNN模型
	knn->setDefaultK(k);//設(shè)定k值
	knn->setIsClassifier(true);//KNN算法可用于分類、回歸。
	knn->setAlgorithmType(KNearest::BRUTE_FORCE);//字符匹配算法
	knn->train(TrainDataMat, ROW_SAMPLE, TrainLabelMat);//模型訓(xùn)練

	//預(yù)測(cè)及結(jié)果顯示
	double count = 0.0;
	Scalar color;
	for (int i = 0; i < TestData.size(); i++)
	{
		//將測(cè)試圖片轉(zhuǎn)成CV_32FC1，單行形式
		Mat data = TestData[i].mat.reshape(0, 1);
		data.convertTo(data, CV_32FC1);
		Mat sample = data(Range(0, data.rows), Range::all());

		float f = knn->predict(sample); //預(yù)測(cè)
		if (f == TestData[i].label)
		{
			color = Scalar(0, 255, 0); //如果預(yù)測(cè)正確，繪制綠色，并且結(jié)果+1
			count++;
		}
		else
		{
			color = Scalar(0, 0, 255);//如果預(yù)測(cè)錯(cuò)誤，繪制紅色
		}

		rectangle(src, TestData[i].rect, color, 2);
	}

	//將繪制結(jié)果拷貝到一張新圖上
	Mat result(Size(src.cols, src.rows + 50), CV_8UC3, Scalar::all(255));
	src.copyTo(result(Rect(0, 0, src.cols, src.rows)));
	//將得分在結(jié)果圖上顯示
	char text[10];
	int score = (count / testNum) * 100;
	sprintf_s(text, "%s%d%s", "Score:", score, "%");
	putText(result, text, Point((result.cols / 2) - 80, result.rows - 15), FONT_HERSHEY_SIMPLEX, 1, Scalar(0, 255, 0), 2);
	imshow("test", result);
	imwrite("result.jpg", result);
	waitKey(0);
	system("pause");
	return 0;
}