win10，Qt基于国内某大厂OCR识别实时检测系统软件

---

前言

  最近与国内某大厂合作，开发了一套基于其OCR识别算法模型的实时监测系统，主要实现的是通过摄像头实时拍摄显示屏的画面，对拍摄的画面做OCR识别后，格式化输出识别结果。
  该软件最终目标为嵌入式设备，现阶段在win10端开始做验证工作。

---

效果

  OCR识别效果如下图所示。在左侧“Input”窗口内显示该OCR识别具体输入内容，右侧“Output”窗口内显示识别后，并做格式化输出后的显示结果。

---

核心代码

编程思路

  这个显示系统，其实是基于了摄像头camera.cpp代码，具体怎样实现接收摄像头图片、并对图像实现加工处理，需要参照以下博客的内容：
https://blog.csdn.net/wang_chao118/article/details/126125022?spm=1001.2014.3001.5502

  基于该大厂的OCR识别模型、SDK文档，将camera.cpp文件进行改写；

camera.cpp

#include "camera.h"
#include 
#include 

using namespace std;
using namespace cv;


// 获取时间戳
std::time_t get_timestamp()
{
    std::chrono::time_point<std::chrono::system_clock, std::chrono::milliseconds> tp
        = std::chrono::time_point_cast<std::chrono::milliseconds>(std::chrono::system_clock::now());
    auto tmp = std::chrono::duration_cast<std::chrono::milliseconds>(tp.time_since_epoch());
    std::time_t timestamp = tmp.count();
    return timestamp;
}
//获取授权key
void get_license_key(std::string& key, const std::string& key_path)
{

    std::fstream fs;
    fs.open(key_path, std::ios::in | std::ios::binary);

    if (!fs) {
        return;
    }

    char line[256];
    fs.getline(line, 256);
    key = std::string(line);
    fs.close();
}


template <class T>
int getlength(T& array)
{
    return (sizeof(array) / sizeof(array[0]));
}

void Camera::folderTest(string path, Character_SDK* api) {
    _finddata64i32_t fileInfo;
    string p = path + "*";
    intptr_t hFile = _findfirst(p.c_str(), &fileInfo);

    if (hFile == -1) {
        cout << "hFile == -1" << endl;
        return;
    }

    do
    {
        if (fileInfo.size > 0) {
            imageProcess(path + fileInfo.name, api);
        }

    } while (_findnext(hFile, &fileInfo) == 0);

    // c++17 以上可以用此方法
    //for (auto& i : filesystem::directory_iterator(path)) {
    //    imageProcess(i.path().string(), api);
   // }
}

void Camera::imageProcess(string path, Character_SDK* api) {
    // 以下为demo示例
    std::time_t time_begin = get_timestamp();
    cv::Mat mat = cv::imread(path);
    CharacterResponse response;
    bool is_single_word = false;
    int processCode = api->process(&mat, is_single_word, &response);
    printf("process result=%d \n", processCode);
    if (processCode != 0) {
        printf("process error=%d \n", processCode);
        return;
    }
    std::string result_str("123456");

    for (int i = 0; i < response.texts.size(); ++i) {
        if (response.texts[i].empty()) {
            continue;
        }
        result_str = result_str + response.texts[i] + "\t";
        result_str = result_str + to_string(response.text_coords[i][0]) + "\t";
        result_str = result_str + to_string(response.text_coords[i][1]) + "\t";
        result_str = result_str + to_string(response.text_coords[i][2]) + "\t";
        result_str = result_str + to_string(response.text_coords[i][3]) + "\t";
        result_str = result_str + to_string(response.text_coords[i][4]) + "\t";
        result_str = result_str + to_string(response.text_coords[i][5]) + "\t";
        result_str = result_str + to_string(response.text_coords[i][6]) + "\t";
        result_str = result_str + to_string(response.text_coords[i][7]) + "\t";

        if (is_single_word) {
            for (int j = 0; j < response.words[i].word.size(); ++j) {
                result_str = result_str + response.words[i].word[j] + "\t";
                result_str = result_str + to_string(response.words[i].word_coords[j][0]) + "\t";
                result_str = result_str + to_string(response.words[i].word_coords[j][1]) + "\t";
                result_str = result_str + to_string(response.words[i].word_coords[j][2]) + "\t";
                result_str = result_str + to_string(response.words[i].word_coords[j][3]) + "\t";
            }
        }

        result_str += "\n";
    }
    std::cout << "result =" << result_str << std::endl;
    string filename = path.substr(10, path.length());
    string outputname = "result/" + filename + ".txt";
    cout << outputname << "===========" << std::endl;
    ofstream fOut(outputname);
    fOut << result_str.c_str() << std::endl;
    fOut.close();

    std::time_t time_end = get_timestamp();
    std::cout << "time cost is :" << time_end - time_begin << "ms" << std::endl;

}

void Camera::imageProcess(cv::Mat src, Character_SDK* api) {

    std::time_t time_begin = get_timestamp();
    cv::Mat mat = src;
    CharacterResponse response;
    bool is_single_word = false;
    int processCode = api->process(&mat, is_single_word, &response);
    printf("process result=%d \n", processCode);
    if (processCode != 0) {
        printf("process error=%d \n", processCode);
        return;
    }
    std::string result_str("\n");

    for (int i = 0; i < response.texts.size(); ++i) {
        if (response.texts[i].empty()) {
            continue;
        }
        result_str = result_str + response.texts[i] + "\t";
        result_str = result_str + to_string(response.text_coords[i][0]) + "\t";
        result_str = result_str + to_string(response.text_coords[i][1]) + "\t";
        result_str = result_str + to_string(response.text_coords[i][2]) + "\t";
        result_str = result_str + to_string(response.text_coords[i][3]) + "\t";
        result_str = result_str + to_string(response.text_coords[i][4]) + "\t";
        result_str = result_str + to_string(response.text_coords[i][5]) + "\t";
        result_str = result_str + to_string(response.text_coords[i][6]) + "\t";
        result_str = result_str + to_string(response.text_coords[i][7]) + "\t";

        if (is_single_word) {
            for (int j = 0; j < response.words[i].word.size(); ++j) {
                result_str = result_str + response.words[i].word[j] + "\t";
                result_str = result_str + to_string(response.words[i].word_coords[j][0]) + "\t";
                result_str = result_str + to_string(response.words[i].word_coords[j][1]) + "\t";
                result_str = result_str + to_string(response.words[i].word_coords[j][2]) + "\t";
                result_str = result_str + to_string(response.words[i].word_coords[j][3]) + "\t";
            }
        }

        result_str += "\n";
    }

    std::cout << "result =" << result_str << std::endl;
    std::time_t time_end = get_timestamp();
    std::cout << "time cost is :" << time_end - time_begin << "ms" << std::endl;

    out_img = output_image(response);

    print_result = QString::fromStdString(result_str);
}

QImage Camera::output_image(CharacterResponse response)
{
    QImage image(640,480, QImage::Format_ARGB32);
    //填充图片背景,120/250为透明度
    image.fill(qRgba(255, 255, 255, 0));


    //为这个QImage构造一个QPainter
    QPainter painter(&image);
    //设置画刷的组合模式CompositionMode_SourceOut这个模式为目标图像在上。
    //改变组合模式和上面的填充方式可以画出透明的图片。
    painter.setCompositionMode(QPainter::CompositionMode_DestinationOver);


    //改变画笔和字体
    QPen pen = painter.pen();
    pen.setColor( QColor(255, 0, 0));

    QFont font;
    //设置显示字体的大小
    font.setPixelSize(22);


    QFontMetrics fm(font);
    painter.setPen(pen);
    painter.setFont(font);


//    CharacterResponse response=*resp;
    for (int i = 0; i < response.texts.size(); ++i) {
        if (response.texts[i].empty()) {
            continue;
        }
        // 只需要使用左上角坐标
        QPoint q(response.text_coords[i][0],response.text_coords[i][1]);
        painter.drawText(q, QString::fromStdString(response.texts[i]));
     }


    //缩放
  QImage scaledimg=image.scaled(640,480,Qt::KeepAspectRatio);

    return scaledimg;
}

Camera::Camera()
{
    std::cout << "before sdk_init" << std::endl;
    api = new Character_SDK();

    //设置保存sdk日志
    api->set_log_config(0,"",false);

    std::string licenseKeyPath = "C:/Users/118ai/Desktop/Baidu_Ocr_230220/Baidu_ocr/x64/license/license.key";
    std::string key;
    get_license_key(key, licenseKeyPath);
    std::string licenseFile = "C:/Users/118ai/Desktop/Baidu_Ocr_230220/Baidu_ocr/x64/license/license.ini";
    //授权
    int res = api->auth_from_file(key.c_str(), licenseFile.c_str());
    if (res != 0)
    {
        std::cout << "sdk auth_from_file  ret code:" << res << std::endl;
        getchar();
    }

   // model_path为模型文件夹路径，即resources文件夹（里面存的是身份证识别的模型文件）
   // 传空为采用默认路径，若想定置化路径，请填写全局路径如：d:\\resource
    std::string model_path = "C:/Users/118ai/Desktop/Baidu_Ocr_230220/Baidu_ocr/x64/Release/resource";
    res = api->sdk_init(model_path.c_str());
    std::cout << "after sdk_init" << std::endl;
    if (res != 0)
    {
        std::cout << "sdk init result is:" << res << std::endl;
        getchar();
    }

    //单次测试
//    imageProcess("C:/Users/118ai/Desktop/Baidu_Ocr_230220/Baidu_ocr/x64/images/avic.jpg", api);
    //文件夹测试
//    folderTest("C:/Users/118ai/Desktop/Baidu_Ocr_230220/Baidu_ocr/x64/images/", api);

}

Camera::~Camera()
{
    std::cout << "before delete api" << std::endl;
    delete api;
}

void Camera::camNumber(const int &n)
{
    camera_num = n;
}

void Camera::openCamera()
{
    capture.set(CAP_PROP_FRAME_WIDTH, 640);
    capture.set(CAP_PROP_FRAME_HEIGHT, 480);

//    capture.open(camera_num,cv::CAP_DSHOW);
    capture.open(camera_num);
    if(!capture.isOpened())
    {
        qDebug() << "camer_num:"<<camera_num<<"not open";
        return;
    }
}

void Camera::mainwindowDisplay()
{
//    QThread::sleep(1);   //耗时操作
    capture >> src_image;
    imageProcess(src_image, api);
    QImage img1 = QImage((const unsigned char*)src_image.data,
                         src_image.cols, src_image.rows,
                         QImage::Format_RGB888).rgbSwapped();

    emit sendPicture(img1);
//    emit send_result(print_result);
    emit send_result_picture(out_img);

 }

资源下载

本案例中涉及到的所有代码、工程文件等，到此处https://download.csdn.net/download/wang_chao118/87586281下载。