OpenCV每日函数 WeChat QR 微信二维码检测器

一、概述

微信二维码检测器，用于检测和解析二维码。

微信二维码包括两个基于 CNN 的模型：物体检测模型和超分辨率模型。对象检测模型用于检测带有边界框的二维码。二维码较小时，应用超分辨率模型放大二维码。

二、类参考

1、函数原型

初始化微信二维码。它包括两个模型，它们以 caffe 格式打包。因此，有prototxt和caffe模型（总共四个参数）。

WeChatQRCode (const std::string &detector_prototxt_path="", const std::string &detector_caffe_model_path="", const std::string &super_resolution_prototxt_path="", const std::string &super_resolution_caffe_model_path="")

2、参数详解

detector_prototxt_path	prototxt file path for the detector
detector_caffe_model_path	caffe model file path for the detector
super_resolution_prototxt_path	prototxt file path for the super resolution model
super_resolution_caffe_model_path	caffe file path for the super resolution model

上述四个模型文件下载地址

GitHub - WeChatCV/opencv_3rdparty: OpenCV - 3rdpartyhttps://github.com/WeChatCV/opencv_3rdparty

3、函数原型

检测和解码 QR 码。

detectAndDecode (InputArray img, OutputArrayOfArrays points=noArray())

4、参数详解

img	支持灰度或彩色（BGR）图像。
points	找到的二维码四边形的顶点的可选输出数组。如果找不到，则为空。

三、OpenCV源码

1、源码路径

opencv_contrib\modules\wechat_qrcode\src\wechat_qrcode.cpp

2、源码代码

WeChatQRCode::WeChatQRCode(const String& detector_prototxt_path,
                           const String& detector_caffe_model_path,
                           const String& super_resolution_prototxt_path,
                           const String& super_resolution_caffe_model_path) {
    p = makePtr<WeChatQRCode::Impl>();
    if (!detector_caffe_model_path.empty() && !detector_prototxt_path.empty()) {
        // initialize detector model (caffe)
        p->use_nn_detector_ = true;
        CV_Assert(utils::fs::exists(detector_prototxt_path));
        CV_Assert(utils::fs::exists(detector_caffe_model_path));
        p->detector_ = make_shared<SSDDetector>();
        auto ret = p->detector_->init(detector_prototxt_path, detector_caffe_model_path);
        CV_Assert(ret == 0);
    } else {
        p->use_nn_detector_ = false;
        p->detector_ = NULL;
    }
    // initialize super_resolution_model
    // it could also support non model weights by cubic resizing
    // so, we initialize it first.
    p->super_resolution_model_ = make_shared<SuperScale>();
    if (!super_resolution_prototxt_path.empty() && !super_resolution_caffe_model_path.empty()) {
        p->use_nn_sr_ = true;
        // initialize dnn model (caffe format)
        CV_Assert(utils::fs::exists(super_resolution_prototxt_path));
        CV_Assert(utils::fs::exists(super_resolution_caffe_model_path));
        auto ret = p->super_resolution_model_->init(super_resolution_prototxt_path,
                                                    super_resolution_caffe_model_path);
        CV_Assert(ret == 0);
    } else {
        p->use_nn_sr_ = false;
    }
}

vector<string> WeChatQRCode::detectAndDecode(InputArray img, OutputArrayOfArrays points) {
    CV_Assert(!img.empty());
    CV_CheckDepthEQ(img.depth(), CV_8U, "");

    if (img.cols() <= 20 || img.rows() <= 20) {
        return vector<string>();  // image data is not enough for providing reliable results
    }
    Mat input_img;
    int incn = img.channels();
    CV_Check(incn, incn == 1 || incn == 3 || incn == 4, "");
    if (incn == 3 || incn == 4) {
        cvtColor(img, input_img, COLOR_BGR2GRAY);
    } else {
        input_img = img.getMat();
    }
    auto candidate_points = p->detect(input_img);
    auto res_points = vector<Mat>();
    auto ret = p->decode(input_img, candidate_points, res_points);
    // opencv type convert
    vector<Mat> tmp_points;
    if (points.needed()) {
        for (size_t i = 0; i < res_points.size(); i++) {
            Mat tmp_point;
            tmp_points.push_back(tmp_point);
            res_points[i].convertTo(((OutputArray)tmp_points[i]), CV_32FC2);
        }
        points.createSameSize(tmp_points, CV_32FC2);
        points.assign(tmp_points);
    }
    return ret;
};

四、效果图像示例

使用了和QRCodeDetector模块相同的图像对微信二维码进行了简单测试。

OpenCV每日函数 Object Detection目标检测模块 (3) 二维码检测和生成_坐望云起的博客-CSDN博客OpenCV提供的二维码生成的类。 OpenCV提供的二维码检测的类。https://skydance.blog.csdn.net/article/details/125300314 参考代码如下

Bitmap bitmap = this.appWorkspace.ActiveDocumentWorkspace.CompositionSurface.CreateAliasedBitmap();
                Mat src = OpenCvSharp.Extensions.BitmapConverter.ToMat(bitmap);

                OpenCvSharp.WeChatQRCode weChatQRCode = WeChatQRCode.Create(
                    "wechat\\detect.prototxt",
                    "wechat\\detect.caffemodel",
                    "wechat\\sr.prototxt",
                    "wechat\\sr.caffemodel");

                Mat[] mats;
                string[] res;
                weChatQRCode.DetectAndDecode(src, out mats, out res);

                if(res!=null && res.Length > 0)
                {
                    foreach (string s in res)
                        this.richTextBox1.Text += s;
                }

示例图像如下