手写数字识别，C++，OpenCV

图片切割：

// 核心代码：
char root_path[150];    // 数字图片的路径
char pic_path[100];
sprintf(root_path, "E:\4_大三下\机器视觉\实验图片\8\8-1.bmp");
IplImage *inputImg = cvLoadImage(root_path,1);
int block = 4;
int nWidth = inputImg->width / block;
int nHeight = inputImg->height / block;
for(int i = 0;i < block;i++)
    for(int j = 0; j < block;j++) {
        cvResetImageROI(inputImg);
        cvSetImageROI(inputImg, cvRect(i * nWidth, j * nHeight, nWidth, nHeight));
        IplImage * img = cvCreateImage(cvSize(nWidth, nHeight), 8, 3);
        cvCopy(inputImg,img);
        sprintf(pic_path, "E:\4_大三下\机器视觉\%d-%d.bmp", i, j);
        cvSaveImage(pic_path, img);
    }

实现数字图像的特征提取算法，求0~9共十个数字的平均特征向量；

实现数字图像的识别算法，任意输入三个不同的数字图像，将数字图像和识别结果输出到屏幕；

#include "stdio.h"
#include "math.h"
#include "vector"
#include "cv.h"
#include "highgui.h"

using namespace std;

// 函数功能，计算一张图片某一个矩形块(以x,y为原点， width,height为宽度和高度)中黑色像素点的个数
int ImageBlackCount(IplImage *Img, int x,int y,int width,int height)
{
    int nCount = 0;
    char *data = (char *)Img->imageData;    // 新图象数据信息
    int  wp = Img->widthStep;
    for(int i = 0;i < height;i++)
        for(int j = 0;j < width;j++)
        {
            int r = data[(x + i) * wp + 3 * (y + j)];
            if(r == 0)
                nCount++;
        }
    return nCount;
}

// 函数功能，返回一张图片的特征向量
vector<int> ImageVec(IplImage *Img)
{
    vector<int> temp(16,0);
    int nwidth = Img->width / 4;
    int nheight = Img->height / 4;        // 分成16块
    int x = 0, y = 0;
    for(int k = 1;k <= 16;k++) 
    {
        temp[k - 1] = ImageBlackCount(Img, x, y, nwidth, nheight);
        y += nwidth;
        if(k % 4 == 0)
        {
            x += nheight;
            y = 0;
        }
    }    
    return temp;
}

// 计算欧氏距离
double CalDist(vector<int> test,vector<int> train)
{
    double dist = 0;
    if(test.size() == train.size())
        for(int i = 0;i < test.size();i++)
            dist += (train[i] - test[i]) * (train[i] - test[i]);
    return sqrt(dist);
}

int main(int arge,char * argv[])
{
    char root_path[150];    // 数字图片的路径
    vector< vector<int> > vec;
    // 模板库--平均特征向量求解
    int block = 0;
    for(int i = 0;i < 10;i++)
    {
        vector<int> avg(16,0);
        for(int j = block +1; j <= block + 5;j++) {
            sprintf(root_path, "E:\4_大三下\机器视觉\实验图片\%d\%d-%d.bmp", i, i, j);
            IplImage *inputImg = cvLoadImage(root_path,1);
            vector<int> temp = ImageVec(inputImg);
            for(int k = 0;k < 16;k++)
                avg[k] += temp[k];
        }
        printf("数字%d的平均特征向量：",i);
        for(int k = 0;k < 16;k++) {
            avg[k] = avg[k] / 5;//block;
            printf("%d ",avg[k]);        // 一个数字的平均特征向量
        }
        putchar(10);
        vec.push_back(avg);
    }

    // 剩余数字的测试匹配
    char pic_path[150];
    vector< vector<int> > Mat;
    for(int i2 = 0;i2 < 10;i2++)
    {
        vector<int> mat_test(10,0);
        for(int j2 = 6 - block; j2 <= 10 - block;j2++) {
            sprintf(pic_path, "E:\4_大三下\机器视觉\实验图片\%d\%d-%d.bmp", i2, i2, j2);
            IplImage *test = cvLoadImage(pic_path,1);
            vector<int> temp = ImageVec(test);
            int cls = 0;
            double d = 0, min = CalDist(temp, vec[0]);
            for(int n = 0;n < 10;n++) 
            {
                d = CalDist(temp, vec[n]);
                if(min >  d) 
                {
                    min = d;
                    cls = n;
                }
            }
            mat_test[cls]++;
        }
        Mat.push_back(mat_test);
    }
    double rate = 0;
    // 混淆矩阵
    printf("   0 1 2 3 4 5 6 7 8 9 
");
    for(int i3 = 0;i3 < 10;i3++) 
    {
        rate += Mat[i3][i3];
        printf("%d: ",i3);
        for(int j3 = 0; j3 < 10;j3++)
            printf("%d ",Mat[i3][j3]);
        putchar(10);
    }
    printf("总识别率为:%.2lf%%
",rate * 2);
    return 0;
}

 2021-06-04