一、车牌的识别和校正
本文采用一工程多项目模式，以代码呈现，因还未接触MFC，所以敬请见谅，之后会继续学习，不断完善代码。
车牌识别借鉴于CSDN博主吾理小子的博客，表达由衷的感谢！https://blog.csdn.net/qq_39960119/article/details/83930112
对其中的一些参数和定义做了一些修改，增加了对倾斜图片的修正，不过鄙人对于倾斜角度参数的理解依旧不到位，因此对于角度的处理还是不太理解，属实惭愧。
#include <iostream>
#include <opencv2\opencv.hpp>
using namespace std;
using namespace cv;
int main()
{
Mat OriginalImg;
OriginalImg = imread("TestPhoto.jpg", IMREAD_COLOR);//读取原始彩色图像
if (OriginalImg.empty()) //判断图像对否读取成功
{
cout << "错误!读取图像失败\n";
return -1;
}
// imshow("原图", OriginalImg); //显示原始图像
cout << "Width:" << OriginalImg.rows << "\tHeight:" << OriginalImg.cols << endl;//打印长宽
Mat ResizeImg;
//if (OriginalImg.cols > 640)
resize(OriginalImg, ResizeImg, Size(640, 640 * OriginalImg.rows / OriginalImg.cols));
imshow("尺寸变换图", ResizeImg);
unsigned char pixelB, pixelG, pixelR; //记录各通道值
unsigned char DifMax = 65; //基于颜色区分的阈值设置
unsigned char B = 200, G = 80, R = 50; //各通道的阈值设定，针对与蓝色车牌
Mat BinRGBImg = ResizeImg.clone(); //二值化之后的图像
int i = 0, j = 0;
for (i = 0; i < ResizeImg.rows; i++) //通过颜色分量将图片进行二值化处理
{
for (j = 0; j < ResizeImg.cols; j++)
{
pixelB = ResizeImg.at<Vec3b>(i, j)[0]; //获取图片各个通道的值
pixelG = ResizeImg.at<Vec3b>(i, j)[1];
pixelR = ResizeImg.at<Vec3b>(i, j)[2];
if (abs(pixelB - B) < DifMax && abs(pixelG - G) < DifMax && abs(pixelR - R) < DifMax)
{
//将各个通道的值和各个通道阈值进行比较
BinRGBImg.at<Vec3b>(i, j)[0] = 255; //符合颜色阈值范围内的设置成白色
BinRGBImg.at<Vec3b>(i, j)[1] = 255;
BinRGBImg.at<Vec3b>(i, j)[2] = 255;
}
else
{
BinRGBImg.at<Vec3b>(i, j)[0] = 0; //不符合颜色阈值范围内的设置为黑色
BinRGBImg.at<Vec3b>(i, j)[1] = 0;
BinRGBImg.at<Vec3b>(i, j)[2] = 0;
}
}
}
imshow("基于颜色信息二值化", BinRGBImg); //显示二值化处理之后的图像
Mat BinOriImg; //形态学处理结果图像
Mat element = getStructuringElement(MORPH_RECT, Size(3, 3)); //设置形态学处理窗的大小
dilate(BinRGBImg, BinOriImg, element,Point(-1, -1), 5); //进行多次膨胀操作
erode(BinOriImg, BinOriImg, element,Point(-1,-1),5); //进行多次腐蚀操作
imshow("形态学处理后", BinOriImg); //显示形态学处理之后的图像
//--------------------------------------------------------------------------
double length, area, rectArea; //定义轮廓周长、面积、外界矩形面积
double rectDegree = 0.0; //矩形度=外界矩形面积/轮廓面积，比值越大说明效果越好
double long2Short = 0.0; //体态比=长边/短边
CvRect rect; //外界矩形：结构体包含x，y坐标，width和height
CvBox2D box, boxTemp; //外接矩形
CvPoint2D32f pt[4]; //矩形定点变量
double axisLong = 0.0, axisShort = 0.0; //矩形的长边和短边
/*double axisLongTemp = 0.0, axisShortTemp = 0.0;*///矩形的长边和短边
double LengthTemp; //中间变量
float angle = 0; //记录车牌的倾斜角度
bool TestPlantFlag = 0; //车牌检测成功标志位
cvtColor(BinOriImg, BinOriImg, CV_BGR2GRAY); //将形态学处理之后的图像转化为灰度图像
threshold(BinOriImg, BinOriImg, 100, 255, THRESH_BINARY); //灰度图像二值化,//OTSU算法（双峰图效果明显）
CvMemStorage *storage = cvCreateMemStorage(0);//跟栈类似
CvSeq * seq = 0; //创建一个序列,CvSeq本身就是一个可以增长的序列，不是固定的序列
CvSeq * tempSeq = cvCreateSeq(CV_SEQ_ELTYPE_POINT, sizeof(CvSeq), sizeof(CvPoint), storage);//以点坐标形式，序列头大小，储存元素大小，储存在之前的容器里
int cnt = cvFindContours(&(IplImage(BinOriImg)), storage, &seq, sizeof(CvContour), CV_RETR_EXTERNAL, CV_CHAIN_APPROX_SIMPLE);
//第一个参数是IplImage指针类型，将MAT强制转换为IplImage指针类型
//返回轮廓的数目
//获取二值图像中轮廓的个数
cout << "number of contours " << cnt << endl; //打印轮廓个数
for (tempSeq = seq; tempSeq != NULL; tempSeq = tempSeq->h_next)
{
length = cvArcLength(tempSeq); //获取轮廓周长
area = cvContourArea(tempSeq); //获取轮廓面积
if (area > 800 && area < 50000) //矩形区域面积大小判断
{
rect = cvBoundingRect(tempSeq, 1);//计算矩形边界
boxTemp = cvMinAreaRect2(tempSeq, 0); //获取轮廓的矩形
cvBoxPoints(boxTemp, pt); //获取矩形四个顶点坐标
angle = boxTemp.angle; //得到车牌倾斜角度
axisLong = sqrt(pow(pt[1].x - pt[0].x, 2) + pow(pt[1].y - pt[0].y, 2)); //计算长轴（勾股定理）
axisShort = sqrt(pow(pt[2].x - pt[1].x, 2) + pow(pt[2].y - pt[1].y, 2)); //计算短轴（勾股定理）
Point2d points;
vector<Point>dots;
for (int i = 0; i < 4; i++)
{
points.x = pt[i].x;
points.y = pt[i].y;
dots.push_back(points);
}
RotatedRect rect = minAreaRect(dots);
if (axisShort > axisLong) //短轴大于长轴，交换数据
{
LengthTemp = axisLong;
axisLong = axisShort;
axisShort = LengthTemp;
}
else
angle += 90;
rectArea = axisLong * axisShort; //计算矩形的面积
rectDegree = area / rectArea; //计算矩形度（比值越接近1说明越接近矩形）
long2Short = axisLong / axisShort; //计算长宽比
if (long2Short > 2.2 && long2Short < 3.8 && rectDegree > 0.63 && rectDegree < 1.37 && rectArea > 2000 && rectArea < 50000)
{
Mat GuiRGBImg = ResizeImg.clone();
TestPlantFlag = true; //检测车牌区域成功
for (int i = 0; i < 4; ++i) //划线框出车牌区域
cvLine(&(IplImage(GuiRGBImg)), cvPointFrom32f(pt[i]), cvPointFrom32f(pt[((i + 1) % 4) ? (i + 1) : 0]), CV_RGB(255, 0, 0));//实现闭口画线
imshow("提取车牌结果图", GuiRGBImg); //显示最终结果图
if (angle != 0)
{
Point2f center(pt->x + (axisLong / 2), pt->y - (axisLong / 2));
Mat warp = getRotationMatrix2D(center, angle, 1.0);
warpAffine(OriginalImg, OriginalImg, warp, Size(640, 640 * OriginalImg.rows / OriginalImg.cols));//不设置会出现内存问题
resize(OriginalImg, OriginalImg, Size(640, 640 * OriginalImg.rows / OriginalImg.cols));
imshow("旋转后的原图", OriginalImg);
imwrite("affineimg.jpg", OriginalImg);
}
else
{
Mat img_ROI = GuiRGBImg(Rect(pt->x+2, pt->y - axisShort+2, axisLong, axisShort));//提取感兴趣区域这里+2是为了修正红色矩形边框
imshow("车牌", img_ROI);
resize(img_ROI, img_ROI, Size(354, 118));
imshow("车牌2", img_ROI);
imwrite("affineimg.jpg", img_ROI);
}
cout << "倾斜角度：" << angle << endl;
}
}
}
waitKey();
return 0;
}
代码运行效果图如下：
二、获取车牌
此处是对修正后的原图进行的处理，提取为354×118像素的车牌校正后的图片
识别与提取的代码上面写过了，拿过来用就行。
Mat img_ROI = GuiRGBImg(Rect(pt->x - axisLongTemp+2, pt->y - axisShortTemp+2, axisLongTemp-5, axisShortTemp-5));//提取感兴趣区域
imshow("车牌", img_ROI);
resize(img_ROI, img_ROI, Size(354, 118));
imshow("车牌2", img_ROI);
imwrite("img_ROI.jpg", img_ROI);
运行结果如下：
二·1 边缘检测法
借鉴于CSDN博主Nine-days的部分代码，并做了一些完善和普适兼容。表达由衷的感谢！https://blog.csdn.net/u011808673/article/details/78510692
int main()
{
Mat OriginalImg;
OriginalImg = imread("blurcar.jpg", IMREAD_COLOR);//读取原始彩色图像
if (OriginalImg.empty()) //判断图像对否读取成功
{
cout << "错误!读取图像失败\n";
return -1;
}
cout << "Width:" << OriginalImg.rows << "\tHeight:" << OriginalImg.cols << endl;//打印长宽
Mat ResizeImg;
resize(OriginalImg, ResizeImg, Size(640, 640 * OriginalImg.rows / OriginalImg.cols));
imshow("尺寸变换图", ResizeImg);
Mat gray_img;
cvtColor(ResizeImg, gray_img, CV_RGB2GRAY);
Mat blur_img;
blur(gray_img, blur_img, Size(3, 3));
Mat candy_img;
Canny(blur_img, candy_img, 300, 100, 3);
imshow("test", candy_img);
//形态学处理
//图片膨胀处理
Mat dilate_image, erode_image, BinOriImg;
//自定义 核进行 x 方向的膨胀腐蚀
Mat elementX = getStructuringElement(MORPH_RECT, Size(22, 1));
Mat elementY = getStructuringElement(MORPH_RECT, Size(1, 20));
Point point(-1, -1);
dilate(candy_img, dilate_image, elementX, point, 2);
erode(dilate_image, erode_image, elementX, point, 4);
dilate(erode_image, dilate_image, elementX, point, 2);
//自定义 核进行 Y 方向的膨胀腐蚀
erode(dilate_image, erode_image, elementY, point, 1);
dilate(erode_image, BinOriImg, elementY, point, 2);
imwrite("dilate_image.jpg", BinOriImg);
//噪声处理
//平滑处理 中值滤波
Mat blur_image;
medianBlur(BinOriImg, blur_image, 15);
medianBlur(blur_image, blur_image, 15);
imshow("test2", blur_image);
接下来就是对处理过的图像进行车牌提取，代码上面写过了，拿来用就行，运行结果如下图：
三、字符分割
此字符分割解决了垂直投影切割“川”字和其他易于被分割错误的汉字的问题，简化了对于车牌中的点被分割的问题。
借鉴于博主lxx_123456的文章，表达由衷的感谢！https://blog.csdn.net/lxx_123456/article/details/79078570
#define _CRT_SECURE_NO_WARNINGS
#define cols_value 0 //cols:2 row:1对于edge搜索
#define row_value 0 //对于川来说不需要去边框
#include <opencv2/opencv.hpp>
#include <math.h>
#include<vector>
#include<cv.h>
using namespace cv;
using namespace std;
vector<Mat> verticalProjectionMat(Mat Image)//封装垂直投影
{
int perPixelValue;//每个像素的值
int width = Image.cols;
int height = Image.rows;
printf("图片的宽%d图片的高%d", width, height);
int* projectValArry = new int[width];//创建用于储存每列白色像素个数的数组
memset(projectValArry, 0, width * 4);//初始化数组
for (int col = 0; col < width; col++)//列
{
int cols_convert_num = 0;
for (int i = 0; i < height - 1; i++)
{
if (Image.at<uchar>(i, col) != Image.at<uchar>(i + 1, col))
cols_convert_num++;
}
if (cols_convert_num < cols_value)
{
continue;
}
for (int row = 0; row < height; row++)//行
{
int row_convert_num = 0;
for (int j = 0; j < width - 1; j++)
{
if (Image.at<uchar>(row,j) != Image.at<uchar>(row,j+1))
row_convert_num++;
}
if (row_convert_num < row_value)
{
continue;
}
perPixelValue = Image.at<uchar>(row, col);//每个像素的值
//if (perPixelValue == 0)//如果是白底黑字
if (perPixelValue == 255)//如果是黑底白字
{
projectValArry[col]++;//列上的叠加
}
}
}
Mat verticalProjectionMat(height, width, CV_8U, Scalar(255));//垂直投影的画布
for (int i = 0; i < height; i++)
{
for (int j = 0; j < width; j++)
{
perPixelValue = 255; //背景设置为白色
verticalProjectionMat.at<uchar>(i, j) = perPixelValue;//遍历设置背景颜色
}
}
for (int i = 0; i < width; i++)//垂直投影直方图
{
for (int j = 0; j < projectValArry[i]; j++)
{
perPixelValue = 0; //直方图设置为黑色
verticalProjectionMat.at<uchar>(height - 1 - j, i) = perPixelValue;
}
}
imshow("垂直投影", verticalProjectionMat);//以上是如何让这个画布形成的呢
Rect rect(0, 0, 120, 40);
Mat image_cut = Mat(verticalProjectionMat, rect);
Mat image_copy = image_cut.clone();
//imshow("切割图片", image_copy);
vector<Mat> roiList;//用于储存分割出来的每个字符
int startIndex = 0;//记录进入字符区的索引
int endIndex = 0;//记录进入空白区域的索引
bool inBlock = false;//是否遍历到了字符区内
for (int i = 0; i < Image.cols; i++)//cols=width
{
if (!inBlock && projectValArry[i] != 0)//进入字符区
{
inBlock = true;
startIndex = i;
}
else if (projectValArry[i] ==0 && inBlock)//进入空白区
{
while (i < Image.cols / 7)//分割汉字
{
i++;
}
endIndex = i;
inBlock = false;
Mat roiImg = Image(Range(0, Image.rows), Range(startIndex, endIndex + 1));
roiList.push_back(roiImg);
}
}
delete[] projectValArry;
return roiList;
}
int main()
{
Point point(-1, -1);
//Mat Image = imread("E:\\LicenseRecognition\\EdgeSearch\\EdgeSearch\\img_ROI.jpg");
Mat Image = imread("E:\\LicenseRecognition\\EdgeSearch\\GetPointedLabel\\img_ROI.jpg");//可用不同的路径
Mat Image1;
cvtColor(Image, Image1, CV_BGR2GRAY);
imshow("灰度化", Image1);
Mat Image2;
threshold(Image1, Image2, 158, 255, CV_THRESH_BINARY);//二值化//100,255 Edge//157,255颜色分割 162edge分割
imshow("二值化", Image2);
Mat Image3;
Mat element = getStructuringElement(MORPH_RECT, Size(1,1));
morphologyEx(Image2, Image3, MORPH_OPEN, element,point,4);//开运算
imshow("开运算", Image3);
int size = 0;
char szName[30] = {
0 };
vector<Mat> b = verticalProjectionMat(Image3);
for (int j = 0; j < b.size(); j++)
{
if (j == 2)//去除车牌中的点
{
continue;
}
sprintf(szName, "vertical_%d.jpg", j);
resize(b[j], b[j],Size(20, 40));//不可调整顺序，不然质量差
imshow(szName, b[j]);
imwrite(szName, b[j]);
}
waitKey(0);
}
分割效果：
四、模板匹配以及识别
借鉴于某车牌识别系统开源源码中模板匹配及识别一小部分。
缺点1：8与B、5与6的识别不够准确（因为是像素相减），幸好鄙人多测试了几张图，发现8与B、5与6的判断标准恰巧是相反的，算作投机取巧，此为学习研究人士的大忌，若无奈之才疏学浅，实不可取。
缺点2：识别的汉字有限，代码后加了其他省份的车牌汉字可用作代码识别修改
如有大佬有解决缺点完善代码之法可以积极评论指正，在下感激不尽！
//此为head.h头文件
#pragma once
#include "cv.h"
#include "highgui.h"
struct pattern
{
double feature[33]; //样本的特征向量
int number; //待识别字符在样本库中的序列号
};
//定义特征提取函数
void GetFeature(IplImage *src, pattern &pat);
#define _CRT_SECURE_NO_WARNINGS
#include "head.h"
#include<opencv2/opencv.hpp>
#include<cstring>
#include<highgui/highgui.hpp>
using namespace cv;
using namespace std;
void GetFeature(IplImage* src, pattern &pat)
{
CvScalar s;
int i, j;
for (i = 0; i < 33; i++)
pat.feature[i] = 0.0;
//图像大小是20*40大小的，分成25块
//********第一行***********
//第一块
for (j = 0; j < 8; j++)
{
for (i = 0; i < 4; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[0] += 1.0;
}
}
//第二块
for (j = 0; j < 8; j++)
{
for (i = 4; i < 8; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[1] += 1.0;
}
}
//第三块
for (j = 0; j < 8; j++)
{
for (i = 8; i < 12; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[2] += 1.0;
}
}
//第四块
for (j = 0; j < 8; j++)
{
for (i = 12; i < 16; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[3] += 1.0;
}
}
//第五块
for (j = 0; j < 8; j++)
{
for (i = 16; i < 20; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[4] += 1.0;
}
}
//********第二行***********
//第六块
for (j = 8; j < 16; j++)
{
for (i = 0; i < 4; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[5] += 1.0;
}
}
//第七块
for (j = 8; j < 16; j++)
{
for (i = 4; i < 8; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[6] += 1.0;
}
}
//第八块
for (j = 8; j < 16; j++)
{
for (i = 8; i < 12; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[7] += 1.0;
}
}
//第九块
for (j = 8; j < 16; j++)
{
for (i = 12; i < 16; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[8] += 1.0;
}
}
//第十块
for (j = 8; j < 16; j++)
{
for (i = 16; i < 20; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[9] += 1.0;
}
}
//********第三行***********
//第十一块
for (j = 16; j < 24; j++)
{
for (i = 0; i < 4; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[10] += 1.0;
}
}
//第十二块
for (j = 16; j < 24; j++)
{
for (i = 4; i < 8; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[11] += 1.0;
}
}
//第十三块
for (j = 16; j < 24; j++)
{
for (i = 8; i < 12; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[12] += 1.0;
}
}
//第十四块
for (j = 16; j < 24; j++)
{
for (i = 12; i < 16; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[13] += 1.0;
}
}
//第十五块
for (j = 16; j < 24; j++)
{
for (i = 16; i < 20; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[14] += 1.0;
}
}
//********第四行***********
//第十六块
for (j = 24; j < 32; j++)
{
for (i = 0; i < 4; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[15] += 1.0;
}
}
//第十七块
for (j = 24; j < 32; j++)
{
for (i = 4; i < 8; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[16] += 1.0;
}
}
//第十八块
for (j = 24; j < 32; j++)
{
for (i = 8; i < 12; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[17] += 1.0;
}
}
//第十九块
for (j = 24; j < 32; j++)
{
for (i = 12; i < 16; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[18] += 1.0;
}
}
//第二十块
for (j = 24; j < 32; j++)
{
for (i = 16; i < 20; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[19] += 1.0;
}
}
//********第五行***********
//第二十一块
for (j = 32; j < 40; j++)
{
for (i = 0; i < 4; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[20] += 1.0;
}
}
//第二十二块
for (j = 32; j < 40; j++)
{
for (i = 4; i < 8; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[21] += 1.0;
}
}
//第二十三块
for (j = 32; j < 40; j++)
{
for (i = 8; i < 12; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[22] += 1.0;
}
}
//第二十四块
for (j = 32; j < 40; j++)
{
for (i = 12; i < 16; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[23] += 1.0;
}
}
//第二十五块
for (j = 32; j < 40; j++)
{
for (i = 16; i < 20; i++)
{
s = cvGet2D(src, j, i);
if (s.val[0] == 255)
pat.feature[24] += 1.0;
}
}
//下面统计方向交点特征
for (i = 0; i < 20; i++)
{
s = cvGet2D(src, 8, i);
if (s.val[0] == 255)
pat.feature[25] += 1.0;
}
for (i = 0; i < 20; i++)
{
s = cvGet2D(src, 16, i);
if (s.val[0] == 255)
pat.feature[26] += 1.0;
}
for (i = 0; i < 20; i++)
{
s = cvGet2D(src, 24, i);
if (s.val[0] == 255)
pat.feature[27] += 1.0;
}
for (i = 0; i < 20; i++)
{
s = cvGet2D(src, 32, i);
if (s.val[0] == 255)
pat.feature[28] += 1.0;
}
for (j = 0; j < 40; j++)
{
s = cvGet2D(src, j, 4);
if (s.val[0] == 255)
pat.feature[29] += 1.0;
}
for (j = 0; j < 40; j++)
{
s = cvGet2D(src, j, 8);
if (s.val[0] == 255)
pat.feature[30] += 1.0;
}
for (j = 0; j < 40; j++)
{
s = cvGet2D(src, j, 12);
if (s.val[0] == 255)
pat.feature[31] += 1.0;
}
for (j = 0; j < 40; j++)
{
s = cvGet2D(src, j, 16);
if (s.val[0] == 255)
pat.feature[32] += 1.0;
}
}
int main()
{
IplImage * dst_image[7];
dst_image[0]= cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterSeperate\\vertical_0.jpg", 0);
dst_image[1] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterSeperate\\vertical_1.jpg", 0);
dst_image[2] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterSeperate\\vertical_3.jpg", 0);
dst_image[3] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterSeperate\\vertical_4.jpg", 0);
dst_image[4] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterSeperate\\vertical_5.jpg", 0);
dst_image[5] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterSeperate\\vertical_6.jpg", 0);
dst_image[6] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterSeperate\\vertical_7.jpg", 0);
IplImage * char_sample[34];//字符样本图像数组
IplImage * hanzi_sample[9];//汉字样本图像数组
pattern char_pattern[34];//定义字符样品库结构数组
pattern hanzi_pattern[9];//定义汉字样品库结构数组
pattern TestSample[7];//定义待识别字符结构数组
//载入字符模板
char_sample[0] = cvLoadImage("template\\0.bmp", 0);
char_sample[1] = cvLoadImage("template\\1.bmp", 0);
char_sample[2] = cvLoadImage("template\\2.bmp", 0);
char_sample[3] = cvLoadImage("template\\3.bmp", 0);
char_sample[4] = cvLoadImage("template\\4.bmp", 0);
char_sample[5] = cvLoadImage("template\\5.bmp", 0);
char_sample[6] = cvLoadImage("template\\6.bmp", 0);
char_sample[7] = cvLoadImage("template\\7.bmp", 0);
char_sample[8] = cvLoadImage("template\\8.bmp", 0);
char_sample[9] = cvLoadImage("template\\9.bmp", 0);
char_sample[10] = cvLoadImage("template\\A.bmp", 0);
char_sample[11] = cvLoadImage("template\\B.bmp", 0);
char_sample[12] = cvLoadImage("template\\C.bmp", 0);
char_sample[13] = cvLoadImage("template\\D.bmp", 0);
char_sample[14] = cvLoadImage("template\\E.bmp", 0);
char_sample[15] = cvLoadImage("template\\F.bmp", 0);
char_sample[16] = cvLoadImage("template\\G.bmp", 0);
char_sample[17] = cvLoadImage("template\\H.bmp", 0);
char_sample[18] = cvLoadImage("template\\J.bmp", 0);
char_sample[19] = cvLoadImage("template\\K.bmp", 0);
char_sample[20] = cvLoadImage("template\\L.bmp", 0);
char_sample[21] = cvLoadImage("template\\M.bmp", 0);
char_sample[22] = cvLoadImage("template\\N.bmp", 0);
char_sample[23] = cvLoadImage("template\\P.bmp", 0);
char_sample[24] = cvLoadImage("template\\Q.bmp", 0);
char_sample[25] = cvLoadImage("template\\R.bmp", 0);
char_sample[26] = cvLoadImage("template\\S.bmp", 0);
char_sample[27] = cvLoadImage("template\\T.bmp", 0);
char_sample[28] = cvLoadImage("template\\U.bmp", 0);
char_sample[29] = cvLoadImage("template\\V.bmp", 0);
char_sample[30] = cvLoadImage("template\\W.bmp", 0);
char_sample[31] = cvLoadImage("template\\X.bmp", 0);
char_sample[32] = cvLoadImage("template\\Y.bmp", 0);
char_sample[33] = cvLoadImage("template\\Z.bmp", 0);
//载入汉字模板
hanzi_sample[0] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\川.bmp", 0);
hanzi_sample[1] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\鄂.bmp", 0);
hanzi_sample[2] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\黑.bmp", 0);
hanzi_sample[3] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\京.bmp", 0);
hanzi_sample[4] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\辽.bmp", 0);
hanzi_sample[5] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\琼.bmp", 0);
hanzi_sample[6] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\湘.bmp", 0);
hanzi_sample[7] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\粤.bmp", 0);
hanzi_sample[8] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\浙.bmp", 0);
//hanzi_sample[0] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\川a.bmp", 0);
//hanzi_sample[1] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\鄂a.bmp", 0);
//hanzi_sample[2] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\黑a.bmp", 0);
//hanzi_sample[3] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\京a.bmp", 0);
//hanzi_sample[4] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\辽a.bmp", 0);
//hanzi_sample[5] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\琼a.bmp", 0);
//hanzi_sample[6] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\湘a.bmp", 0);
//hanzi_sample[7] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\粤a.bmp", 0);
//hanzi_sample[8] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\浙a.bmp", 0);
//hanzi_sample[9] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\苏a.bmp", 0);
//hanzi_sample[10] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\藏a.bmp", 0);
//hanzi_sample[11] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\甘a.bmp", 0);
//hanzi_sample[12] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\赣a.bmp", 0);
//hanzi_sample[13] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\桂a.bmp", 0);
//hanzi_sample[14] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\沪a.bmp", 0);
//hanzi_sample[15] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\吉a.bmp", 0);
//hanzi_sample[16] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\冀a.bmp", 0);
//hanzi_sample[17] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\津a.bmp", 0);
//hanzi_sample[18] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\晋a.bmp", 0);
//hanzi_sample[19] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\鲁a.bmp", 0);
//hanzi_sample[20] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\蒙a.bmp", 0);
//hanzi_sample[21] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\闽a.bmp", 0);
//hanzi_sample[22] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\宁a.bmp", 0);
//hanzi_sample[23] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\青a.bmp", 0);
//hanzi_sample[24] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\陕a.bmp", 0);
//hanzi_sample[25] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\皖a.bmp", 0);
//hanzi_sample[26] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\新a.bmp", 0);
//hanzi_sample[27] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\渝a.bmp", 0);
//hanzi_sample[28] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\豫a.bmp", 0);
//hanzi_sample[29] = cvLoadImage("E:\\LicenseRecognition\\EdgeSearch\\CharacterMatching\\template\\云a.bmp", 0);
//提取字符样本特征
for (int i = 0; i < 34; i++)
{
GetFeature(char_sample[i], char_pattern[i]);
}
//提取汉字字符特征
for (int i = 0; i < 9; i++)
{
GetFeature(hanzi_sample[i], hanzi_pattern[i]);
}
//提取待识别字符特征
for (int i = 0; i < 7; i++)
{
GetFeature(dst_image[i], TestSample[i]);
}
//进行模板匹配
double min = 100000.0;
for (int num = 0; num < 1; num++)
{
for (int i = 0; i < 9; i++)
{
double diff = 0.0;
for (int j = 0; j < 25; j++)
{
diff += fabs(TestSample[num].feature[j] - hanzi_pattern[i].feature[j]);
}
for (int j = 25; j < 33; j++)
{
diff += fabs(TestSample[num].feature[j] - hanzi_pattern[i].feature[j]) * 9;
}
if (diff < min)
{
min = diff;
TestSample[num].number = i;
}
}
}
for (int num = 1; num < 7; num++)
{
double min_min = 1000000.0;
for (int i = 0; i < 34; i++)
{
double diff_diff = 0.0;
for (int j = 0; j < 25; j++)
{
diff_diff += fabs(TestSample[num].feature[j] - char_pattern[i].feature[j]);
}
for (int j = 25; j < 33; j++)
{
diff_diff += fabs(TestSample[num].feature[j] - char_pattern[i].feature[j]);
}
if (diff_diff < min_min)
{
min_min = diff_diff;
TestSample[num].number = i;
}
}
}
String result = "";//存放识别出的字符
for (int i = 0; i < 1; i++)
{
switch (TestSample[i].number)
{
case 0:
result += "川";
break;
case 1:
result += "鄂";
break;
case 2:
result += "黑";
break;
case 3:
result += "京";
break;
case 4:
result += "辽";
break;
case 5:
result += "琼";
break;
case 6:
result += "湘";
break;
case 7:
result += "粤";
break;
case 8:
result += "浙";
break;
default:
cout<<("识别失败")<<endl;
break;
}
}
for (int i = 1; i < 7; i++)
{
switch (TestSample[i].number)
{
case 0:
result += "0";
break;
case 1:
result += "1";
break;
case 2:
result += "2";
break;
case 3:
result += "3";
break;
case 4:
result += "4";
break;
case 5:
result += "6";
break;
case 6:
result += "5";
break;
case 7:
result += "7";
break;
case 8:
result += "B";
break;
case 9:
result += "9";
break;
case 10:
result += "A";
break;
case 11:
result += "8";
break;
case 12:
result += "C";
break;
case 13:
result += "D";
break;
case 14:
result += "E";;
break;
case 15:
result += "F";
case 16:
result += "G";
break;
case 17:
result += "H";
break;
case 18:
result += "J";
break;
case 19:
result += "K";
break;
case 20:
result += "L";
break;
case 21:
result += "M";
break;
case 22:
result += "N";
break;
case 23:
result += "P";
break;
case 24:
result += "Q";
break;
case 25:
result += "R";
break;
case 26:
result += "S";
break;
case 27:
result += "T";
break;
case 28:
result += "U";
break;
case 29:
result += "U";
break;
case 30:
result += "W";
break;
case 31:
result += "X";
break;
case 32:
result += "Y";
break;
case 33:
result += "Z";
break;
default:
cout<<("识别失败");
break;
}
}
cout<<"车牌的最终结果为："<<result<<endl;//显示结果
system("pause");
return 0;
}
需要template模板，网上下载即可
结果效果：
五、总结
特别感谢CSDN博主吾理小子、CSDN博主Nine-days、CSDN博主lxx_123456等
也感谢CSDN全体制作OpenCV车牌识别有关博客的博主，为此博客奠定了知识基础，此博客仅供学习使用，希望可以给着急于做出车牌识别的朋友一点灵感，不足之处可以指出，如对读者朋友们有用，希望可以点个赞。
六、感悟
这是我第一次通过观摩借鉴复用各位CSDN大佬们的博客文章以及开源代码整合而成的项目，算是我代码路上一个开始，自此逐渐摆脱了拘泥于书本知识而非亲身实践的学习恶习。搞车牌识别项目的时候会遇到很多很多困难与疑惑，通过找博客和参考其他博主的经验和教训来解决自己的困难，比起我之前遇到困难就退缩，只想白嫖成果要好千万倍（虽然这个也白嫖了很多）。希望可以和CSDN上的兄弟们一起进步共同加油！