OPENCV例子opencv-4.5.5\samples\dnn\dasiamrpn_tracker.cpp的代码分析

视觉目标跟踪算法:

• 基于SiamRPN，主要是提出更好的使用数据，利用更好的训练方式让tracker变得更鲁邦
• 有了更好的分数作为指导后，算法可以扩展到Long-term

示例run函数调用情况如下：

 

示例run函数流程图情况如下：

 

示例run函数UML逻辑图情况如下：

 

示例源代码如下：

// DaSiamRPN tracker.

// Original paper: https://arxiv.org/abs/1808.06048

// Link to original repo: https://github.com/foolwood/DaSiamRPN

// Links to onnx models:

// - network:     https://www.dropbox.com/s/rr1lk9355vzolqv/dasiamrpn_model.onnx?dl=0

// - kernel_r1:   https://www.dropbox.com/s/999cqx5zrfi7w4p/dasiamrpn_kernel_r1.onnx?dl=0

// - kernel_cls1: https://www.dropbox.com/s/qvmtszx5h339a0w/dasiamrpn_kernel_cls1.onnx?dl=0

#include <iostream>

#include <cmath>

#include <opencv2/dnn.hpp>

#include <opencv2/imgproc.hpp>

#include <opencv2/highgui.hpp>

#include <opencv2/video.hpp>

using namespace cv;

using namespace cv::dnn;

const char *keys =

        "{ help     h  |   | Print help message }"

        "{ input    i  |   | Full path to input video folder, the specific camera index. (empty for camera 0) }"

        "{ net         | dasiamrpn_model.onnx | Path to onnx model of net}"

        "{ kernel_cls1 | c'c_kernel_cls1.onnx | Path to onnx model of kernel_r1 }"

        "{ kernel_r1   | dasiamrpn_kernel_r1.onnx | Path to onnx model of kernel_cls1 }"

        "{ backend     | 0 | Choose one of computation backends: "选择一个计算后端

                            "0: automatically (by default), "

                            "1: Halide language (http://halide-lang.org/), "卤化物语言

                            "2: Intel's Deep Learning Inference Engine英特尔的深度学习推理引擎(https://software.intel.com/openvino-toolkit), "

                            "3: OpenCV implementation, "

                            "4: VKCOM, "

                            "5: CUDA },"

        "{ target      | 0 | Choose one of target computation devices: "选择目标计算设备之一

                            "0: CPU target (by default), "

                            "1: OpenCL, "

                            "2: OpenCL fp16 (half-float precision), "

                            "3: VPU, "

                            "4: Vulkan, "

                            "6: CUDA, "

                            "7: CUDA fp16 (half-float preprocess) }"

;

static

int run(int argc, char** argv)

{

    // Parse command line arguments. 解析命令行参数

    CommandLineParser parser(argc, argv, keys);

    if (parser.has("help"))

    {

        parser.printMessage();

        return 0;

    }

    std::string inputName = parser.get<String>("input");

    std::string net = parser.get<String>("net");

    std::string kernel_cls1 = parser.get<String>("kernel_cls1");

    std::string kernel_r1 = parser.get<String>("kernel_r1");

    int backend = parser.get<int>("backend");

    int target = parser.get<int>("target");

    Ptr<TrackerDaSiamRPN> tracker;

    try

    {

        TrackerDaSiamRPN::Params params;

        params.model = samples::findFile(net);

        params.kernel_cls1 = samples::findFile(kernel_cls1);

        params.kernel_r1 = samples::findFile(kernel_r1);

        params.backend = backend;

        params.target = target;

        tracker = TrackerDaSiamRPN::create(params);

    }

    catch (const cv::Exception& ee)

    {

        std::cerr << "Exception: " << ee.what() << std::endl;

        std::cout << "Can't load the network by using the following files:" << std::endl;

        std::cout << "siamRPN : " << net << std::endl;

        std::cout << "siamKernelCL1 : " << kernel_cls1 << std::endl;

        std::cout << "siamKernelR1 : " << kernel_r1 << std::endl;

        return 2;

    }

    const std::string winName = "DaSiamRPN";

    namedWindow(winName, WINDOW_AUTOSIZE);

    // Open a video file or an image file or a camera stream. 打开视频文件或图像文件或摄像头流

    VideoCapture cap;

    if (inputName.empty() || (isdigit(inputName[0]) && inputName.size() == 1))

    {

        int c = inputName.empty() ? 0 : inputName[0] - '0';

        std::cout << "Trying to open camera #" << c << " ..." << std::endl;

        if (!cap.open(c))

        {

            std::cout << "Capture from camera #" << c << " didn't work. Specify -i=<video> parameter to read from video file" << std::endl;

            return 2;

        }

    }

    else if (inputName.size())

    {

        inputName = samples::findFileOrKeep(inputName);

        if (!cap.open(inputName))

        {

            std::cout << "Could not open: " << inputName << std::endl;

            return 2;

        }

    }

    // Read the first image.

    Mat image;

    cap >> image;

    if (image.empty())

    {

        std::cerr << "Can't capture frame!" << std::endl;

        return 2;

    }

    Mat image_select = image.clone();

putText(image_select, "Select initial bounding box you want to track.", 选择要跟踪的初始边界框

Point(0, 15), FONT_HERSHEY_SIMPLEX, 0.5, Scalar(0, 255, 0));

    putText(image_select, "And Press the ENTER key.", Point(0, 35), FONT_HERSHEY_SIMPLEX, 0.5, Scalar(0, 255, 0)); 并按 ENTER 键

    Rect selectRect = selectROI(winName, image_select);

    std::cout << "ROI=" << selectRect << std::endl;

    tracker->init(image, selectRect);

    TickMeter tickMeter;

    for (int count = 0; ; ++count)

    {

        cap >> image;

        if (image.empty())

        {

            std::cerr << "Can't capture frame " << count << ". End of video stream?" << std::endl;

            break;

        }

        Rect rect;

        tickMeter.start();

        bool ok = tracker->update(image, rect);

        tickMeter.stop();

        float score = tracker->getTrackingScore();

        std::cout << "frame " << count <<

            ": predicted score=" << score <<

            "  rect=" << rect <<

            "  time=" << tickMeter.getTimeMilli() << "ms" <<

            std::endl;

        Mat render_image = image.clone();

        if (ok)

        {

            rectangle(render_image, rect, Scalar(0, 255, 0), 2);

            std::string timeLabel = format("Inference time: %.2f ms", tickMeter.getTimeMilli());

            std::string scoreLabel = format("Score: %f", score);

            putText(render_image, timeLabel, Point(0, 15), FONT_HERSHEY_SIMPLEX, 0.5, Scalar(0, 255, 0));

            putText(render_image, scoreLabel, Point(0, 35), FONT_HERSHEY_SIMPLEX, 0.5, Scalar(0, 255, 0));

        }

        imshow(winName, render_image);

        tickMeter.reset();

        int c = waitKey(1);

        if (c == 27 /*ESC*/)

            break;

    }

    std::cout << "Exit" << std::endl;

    return 0;

}

int main(int argc, char **argv)

{

    try

    {

        return run(argc, argv);

    }

    catch (const std::exception& e)

    {

        std::cerr << "FATAL: C++ exception: " << e.what() << std::endl;

        return 1;

    }

}