【基于tesseract或ANN的神经网络的身份证号OCR识别】
之前写了一篇,结果浏览器崩了,文字全无。这次直接上代码吧。
身份证号的识别过程:
#include<iostream>
#include<opencv2\opencv.hpp>
using namespace std;
using namespace cv;
int main()
{
Mat img = imread("234.png");
namedWindow("原图");
imshow("原图", img);
Mat temp, temp2, temp3;
//灰度化
cvtColor(img, temp, COLOR_BGR2GRAY);
cvtColor(temp, img, COLOR_GRAY2BGRA);
namedWindow("灰度化");
imshow("灰度化", temp);
//二值化
threshold(temp, temp2, 60, 255, CV_THRESH_BINARY);
namedWindow("二值化");
imshow("二值化", temp2);
//腐蚀
Mat erodeElement = getStructuringElement(MORPH_RECT, Size(13, 13));
erode(temp2, temp3, erodeElement);
namedWindow("腐蚀");
imshow("腐蚀", temp3);
//轮廓检测
vector<vector<Point> > contours; //定义一个容器来存储所有检测到的轮廊
vector<Vec4i> hierarchy;
//轮廓检测函数
Mat conv(temp3.size(), CV_8UC1);
temp3.convertTo(conv, CV_8UC1);
findContours(conv, contours, hierarchy, CV_RETR_TREE, CV_CHAIN_APPROX_NONE, cvPoint(0, 0));
int index = 0;
long max_Index = static_cast<long>(contours.size());
if (index < max_Index - 1) {
++index;
}
//取出身份证号码区域
vector<Rect> rects;
Rect numberRect = Rect(0, 0, 0, 0);
vector<vector<Point> >::const_iterator itContours = contours.begin();
for (int i = 0; i <max_Index; ++i) {
Rect rect = boundingRect(itContours[i]);
numberRect = rect;
Mat dst(numberRect.height, numberRect.width, CV_8UC4, numberRect.area());
dst = img(numberRect);
for (int i = 0; i <max_Index; ++i) {
Rect rect = boundingRect(itContours[i]);
numberRect = rect;
Mat dst(numberRect.height, numberRect.width, CV_8UC4, numberRect.area());
dst = img(numberRect);
if (rect.y > img.rows / 2 && rect.width / rect.height > 6) {
imshow("身份证号:", dst);
}
}
}
waitKey(0);
cout << CV_MAJOR_VERSION << endl;
return 0;
}
运行结果:
然后是基于hp开源谷歌维护的的ocr库,基于vs2015变异的库链接如下:https://download.csdn.net/download/qq_35054151/10765151
第二种方法是根据识别出来的身份证号矩形图,根据垂直投影进行字符分割,然后送入到训练好的数据集中进行识别。
首先,定义一个数组用来储存每一列像素中白色像素的个数。
int perPixelValue;//每个像素的值
int* projectValArry = new int[width];//创建一个用于储存每列白色像素个数的数组
memset(projectValArry, 0, width*4);//必须初始化数组
然后,遍历二值化后的图片,将每一列中白色的(也就是数字区域)像素记录在数组中。
//遍历每一列的图像灰度值,查找每一行255的值
for (int col = 0; col < width; ++col)
{
for (int row = 0; row < height; ++row)
{
perPixelValue = binImg.at<uchar>(row, col);
if (perPixelValue == 255)//如果是黑底白字
{
projectValArry[col]++;
}
}
}
最后,根据数组里的灰度值画出投影图
/*新建一个Mat用于储存投影直方图并将背景置为白色*/
Mat verticalProjectionMat(height, width, CV_8UC1);
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;
}
}
/*将直方图的曲线设为黑色*/
for (int i = 0; i < width; i++)
{
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("【投影】",verticalProjectionMat);
delete[] projectValArry;//不要忘了删除数组空间
有了投影图做切割就很容易了,其实最主要的就是那个储存灰度值的数组,下面就需要根据这个数组的内容来找到相邻字符间的分割点。
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 < srcImg.cols; ++i)
{
if (!inBlock && projectValArry[i] != 0)//进入字符区了
{
inBlock = true;
startIndex = i;
cout << "startIndex is " << startIndex << endl;
}
else if (projectValArry[i] == 0 && inBlock)//进入空白区了
{
endIndex = i;
inBlock = false;
Mat roiImg = srcImg(Range(0,srcImg.rows),Range(startIndex,endIndex+1));
roiList.push_back(roiImg);
}
}
对于分割后的字符的识别,可以参考博主博客:https://blog.csdn.net/qq_35054151/article/details/83685461
另外对于tesseract的识别,可以参考如下:
#include <baseapi.h>
#include <renderer.h>
#include <opencv2\highgui.hpp>
#include <opencv2\imgproc.hpp>
#include <iostream>
void main() {
cv::Mat image = cv::imread("身份证号.jpg");
cv::cvtColor(image, image, CV_BGR2GRAY);
cv::imshow("原灰度图", image);
// eng.traineddata所在路径
//const char* datapath = "D:\\tesseract\\tessdata";
// 新建tess基类
tesseract::TessBaseAPI tess;
// 初始化
tess.Init(NULL, "eng", tesseract::OEM_DEFAULT);
// 设置白名单
tess.SetVariable("tessedit_char_whitelist", "0123456789");
// 设置识别模式
tess.SetPageSegMode(tesseract::PSM_SINGLE_BLOCK);
// 设置识别图像
tess.SetImage((uchar*)image.data, image.cols,
image.rows, image.step[1], image.step[0]);
// 进行识别
char* out = tess.GetUTF8Text();
std::cout << out << std::endl;
cv::waitKey();
}
#include <tesseract/baseapi.h>
tesseract::TessBaseAPI tessearct_api;
const char *languagePath = "/usr/local/Cellar/tesseract/3.04.01_2/share/tessdata"; const char *languageType = "chi_sim";
int nRet = tessearct_api.Init(languagePath, languageType,tesseract::OEM_DEFAULT);
if (nRet != 0) {
printf("初始化字库失败!");
return -1;
}
tessearct_api.SetPageSegMode(tesseract::PSM_SINGLE_BLOCK);
tessearct_api.SetImage(seg_image.data, seg_image.cols, seg_image.rows, 1,seg_image.cols);
string out = string(tessearct_api.GetUTF8Text());
cout<<"the out result :"<<out<<endl;
参考博客:https://blog.csdn.net/wx7788250/article/details/60139109
https://www.jianshu.com/p/3a5c08a14ddd
https://blog.csdn.net/u014389362/article/details/81390350?utm_source=blogxgwz8
更多推荐
所有评论(0)