基于KNN算法实现手写数字识别完整源码+数据（课程设计）.zip

#include<opencv2/opencv.hpp> #include<windows.h> using namespace cv; using namespace std; vector <Mat> result; void setColour(int x) { HANDLE h = GetStdHandle(-11); SetConsoleTextAttribute(h, x); } int main() { int right = 0; int num[10] = { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 }; FileStorage fsClassifications("G:/桌面/数字识别/训练结果/classifications.xml", FileStorage::READ); // 读取 classifications.xml 分类文件 if (fsClassifications.isOpened() == false) { cout << "ERROR, 无法打开classifications.xml\n\n"; system("pause"); return 0; } Mat matClassificationInts; fsClassifications["classifications"] >> matClassificationInts; // 把 classifications.xml 中的 classifications 读取进Mat变量 fsClassifications.release(); // 关闭文件 FileStorage fsTrainingImages("G:/桌面/数字识别/训练结果/images.xml", FileStorage::READ); // 打开训练图片文件 if (fsTrainingImages.isOpened() == false) { cout << "ERROR, 无法打开images.xml\n\n"; system("pause"); return 0; } // 读取训练图片数据（从images.xml中） Mat matTrainingImagesAsFlattenedFloats; fsTrainingImages["images"] >> matTrainingImagesAsFlattenedFloats; // 把 images.xml 中的 images 读取进Mat变量 fsTrainingImages.release(); // 训练 Ptr<ml::KNearest> kNearest(ml::KNearest::create()); // 实例化 KNN 对象 kNearest->setDefaultK(9); // 最终调用train函数，注意到两个参数都是Mat类型（单个Mat） kNearest->train(matTrainingImagesAsFlattenedFloats, ml::ROW_SAMPLE, matClassificationInts); int j = 0; for (int i = 0; i < 100; i++) { char s[100]; sprintf_s(s, "G:/桌面/数字识别/测试集/%d.png", i); Mat inputMat = imread(s, 0); Mat matHandWriting; inputMat.convertTo(matHandWriting, CV_8UC1); Mat matROICopy = matHandWriting.clone(); vector<vector<Point>> contours; vector<Vec4i> hierarchy; findContours(matHandWriting, contours, hierarchy, RETR_EXTERNAL, CHAIN_APPROX_SIMPLE); Rect boundRect = boundingRect(contours[0]); Mat numImage = matROICopy(boundRect).clone(); resize(numImage, numImage, Size(40, 40)); /* 这里计算目标图像的HOG特征（方向梯度直方图特征）代替之前用的灰度特征 */ HOGDescriptor* hog = new HOGDescriptor(Size(40, 40), Size(16, 16), Size(8, 8), Size(8, 8), 9); vector<float> descriptors; hog->compute(numImage, descriptors); Mat matROIFlattenedFloat(1, (int)(descriptors.size()), CV_32FC1, descriptors.data()); Mat matCurrentChar(0, 0, CV_32F); // findNearest的结果保存在这里 // 最终调用 findNearest 函数 kNearest->findNearest(matROIFlattenedFloat, 1, matCurrentChar); //float fltCurrentChar = (float)matCurrentChar.at<float>(0, 0); int intCurrentChar = (int)matCurrentChar.at<float>(0, 0); if (j > 9)j = j % 10; if (num[j] == intCurrentChar) { right++; setColour(7); } else setColour(4); j++; cout << "第"<<i<<"个字符为：" << intCurrentChar << endl; } setColour(15); cout << "正确数 = " << right<<endl; cout << "错误数 = " << 100 - right<<endl; cout << "准确率 = " << (float)right << "%"; int key = waitKey(0); }