test.cpp

#include <iostream>
#include <opencv2/features2d.hpp>
#include <opencv2/opencv.hpp>
#include <map>

// 初始化HSV阈值的上下限
int lowH = 0, lowS = 0, lowV = 0;
int highH = 180, highS = 255, highV = 255;

#define HEIGHT 480
#define EPSILON_RATIO 0.05 //目前这个参数表现良好
#define MIN_AREA_RATIO 0.1  //目前这个参数表现良好
int ROI [] {150,150,300,180};

// 回调函数，用于Trackbar的更新
void on_low_H_thresh_trackbar(int, void *) {}
void on_high_H_thresh_trackbar(int, void *) {}
void on_low_S_thresh_trackbar(int, void *) {}
void on_high_S_thresh_trackbar(int, void *) {}
void on_low_V_thresh_trackbar(int, void *) {}
void on_high_V_thresh_trackbar(int, void *) {}

std::pair<std::vector<cv::Point>, std::vector<cv::Point>> findBox(cv::Mat& img_bin, cv::Mat& frame) {
    if (img_bin.empty()) {
        std::cout << "img_bin cannot be empty" << std::endl;
        throw std::invalid_argument("img_bin cannot be empty");
    }

    int min_area = EPSILON_RATIO * img_bin.total();

    std::map<int, std::vector<cv::Point>> ret;
    std::vector<std::vector<cv::Point>> contours;
    std::vector<cv::Vec4i> hierarchy;

    cv::findContours(img_bin, contours, hierarchy, cv::RETR_TREE, cv::CHAIN_APPROX_SIMPLE);

    if (contours.empty()) {
        return std::make_pair(std::vector<cv::Point>(), std::vector<cv::Point>());
    }

    for (size_t i = 0; i < contours.size(); i++) {
        if (cv::contourArea(contours[i]) >= min_area) {
            double epsilon = EPSILON_RATIO * cv::arcLength(contours[i], true);
            std::vector<cv::Point> approx;
            cv::approxPolyDP(contours[i], approx, epsilon, true);

            if (approx.size() == 4) {
                ret[i] = approx;
            }
        }
    }

    for (size_t i = 0; i < contours.size(); i++) {
        if (ret.count(i) && ret[i].size() == 4 && hierarchy[i][3] != -1 && ret.count(hierarchy[i][3]) && ret[hierarchy[i][3]].size() == 4 && hierarchy[i][2] == -1) {
            if (!frame.empty()) {
                cv::drawContours(frame, contours, i, cv::Scalar(0, 0, 255), 2);
                cv::drawContours(frame, contours, hierarchy[i][3], cv::Scalar(0, 0, 255), 2);
                cv::polylines(frame, ret[i], true, cv::Scalar(0, 255, 255), 10);
                cv::polylines(frame, ret[hierarchy[i][3]], true, cv::Scalar(0, 255, 255), 10);
            }

            return std::make_pair(ret[hierarchy[i][3]], ret[i]);
        }
    }

    return std::make_pair(std::vector<cv::Point>(), std::vector<cv::Point>());
}

int main(int argc, char *argv[])
{
    // 检查是否提供了摄像头路径
    if (argc != 3)
    {
        std::cout << "用法: " << argv[0] << " <摄像头路径> <曝光值>" << std::endl;
        return -1;
    }

    // 打开摄像头
    cv::VideoCapture cap(argv[1], cv::CAP_V4L2);
    if (!cap.isOpened())
    {
        std::cout << "无法打开摄像头: " << argv[1] << std::endl;
        return -1;
    }

    double exposure = std::stod(argv[2]);
    cap.set(cv::CAP_PROP_AUTO_EXPOSURE, 3.0); // 先打开自动曝光
    cap.set(cv::CAP_PROP_AUTO_EXPOSURE, 1.0); // 再关闭自动曝光
    cap.set(cv::CAP_PROP_EXPOSURE, exposure); // 最后设置曝光参数

    // 设置图像的宽度
    bool success1 = cap.set(cv::CAP_PROP_FRAME_WIDTH, 640);
    if (!success1)
    {
        std::cout << "无法设置图像的宽度" << std::endl;
        return -1;
    }

    // 设置图像的高度
    bool success2 = cap.set(cv::CAP_PROP_FRAME_HEIGHT, 480);
    if (!success2)
    {
        std::cout << "无法设置图像的高度" << std::endl;
        return -1;
    }

    // 设置摄像头帧率
    bool success3 = cap.set(cv::CAP_PROP_FPS, 40);
    if (!success3)
    {
        std::cout << "无法设置摄像头帧率" << std::endl;
        return -1;
    }

    // 创建窗口
    cv::namedWindow("Binary Image", cv::WINDOW_AUTOSIZE);

    // 创建Trackbars
    cv::createTrackbar("Low H", "Binary Image", &lowH, 180, on_low_H_thresh_trackbar);
    cv::createTrackbar("High H", "Binary Image", &highH, 180, on_high_H_thresh_trackbar);
    cv::createTrackbar("Low S", "Binary Image", &lowS, 255, on_low_S_thresh_trackbar);
    cv::createTrackbar("High S", "Binary Image", &highS, 255, on_high_S_thresh_trackbar);
    cv::createTrackbar("Low V", "Binary Image", &lowV, 255, on_low_V_thresh_trackbar);
    cv::createTrackbar("High V", "Binary Image", &highV, 255, on_high_V_thresh_trackbar);

    // 设置一些文字的参数
    cv::Point text_position(10, 50); // 在图像的左上角
    int font_face = cv::FONT_HERSHEY_SIMPLEX;
    double font_scale = 1;
    int thickness = 2;

    int MIN_AREA = int(MIN_AREA_RATIO * (ROI[2] - ROI[0]) * (ROI[3] - ROI[1]));

    while (true)
    {
        cv::Mat frame_raw;
        // 读取摄像头的帧
        cap >> frame_raw;
        if (frame_raw.empty())
        {
            std::cout << "无法从摄像头读取帧" << std::endl;
            break;
        }

        // select the ROI region 
        cv::Rect roi(ROI[0], ROI[1], ROI[2], ROI[3]);
        cv::Mat frame = frame_raw(roi);

        // smooth the image
        cv::GaussianBlur(frame, frame, cv::Size(5, 5), 0);

        // 转换到HSV颜色空间
        cv::Mat hsv_image;
        cv::cvtColor(frame, hsv_image, cv::COLOR_BGR2HSV);

        // 定义白色的HSV范围
        cv::Scalar lower_white(lowH, lowS, lowV);
        cv::Scalar upper_white(highH, highS, highV);

        // 应用掩膜以仅保留白色区域
        cv::Mat mask;
        cv::inRange(hsv_image, lower_white, upper_white, mask);

        // 二值化处理
        cv::Mat binary_image;
        cv::bitwise_and(frame, frame, binary_image, mask);

        // 将结果图像转换为二值图像
        cv::cvtColor(binary_image, binary_image, cv::COLOR_BGR2GRAY);
        cv::imshow("grayscale", binary_image);

        // 这个函数的第三个参数thres是需要更改的，对白色的线的检测可以稍微给大一点（滤除多一点假的白点）
        cv::threshold(binary_image, binary_image, 120, 255, cv::THRESH_BINARY);

        auto boxex = findBox(binary_image, frame);

        // draw the contours
        std::vector<std::vector<cv::Point>> contours;
        std::vector<cv::Vec4i> hierarchy;
        cv::findContours(binary_image, contours, hierarchy, cv::RETR_TREE, cv::CHAIN_APPROX_SIMPLE);
        cv::Mat drawing = cv::Mat::zeros(binary_image.size(), CV_8UC3);
        for (size_t i = std::max(0, static_cast<int>(contours.size()) - 2); i < contours.size(); i++)
        {
            cv::Scalar color = cv::Scalar(0, 255, 0);
            cv::drawContours(drawing, contours, (int)i, color, 2, cv::LINE_8, hierarchy, 0);
        }

        cv::imshow("Binary Image", binary_image);
        cv::imshow("Original Image", frame);

        // 按下Esc键退出循环
        char key = (char)cv::waitKey(30);
        if (key == 27)
            break;
    }

    // 释放摄像头
    cap.release();
    cv::destroyAllWindows();

    return 0;
}