调试过程:本次在进行调试过程之中进行了两次垂直检测迭代,进一步去排除水平线的干扰.使用形态学操作去除图片之中的空洞等等.

第一次进行垂直检测,注意这个地方只能够用特定的算子进行垂直检测,别的算子没有这个效果. 

为了减少图片之中白色空洞的干扰,使用开操作.

重复上述操作,进一步排除水平线的干扰.

接下来是进行闭操作,将图中的白色线条尽可能连在一起,上图之中的楼左侧的线有一些断开了.

闭操作的缺陷是会产生小白点点.如下二值化过程

再进行一次反二值化,因为我不会用别的算子结合霍夫直线检测检测出来直线,只能转回去进行操作. 

形态学操作,去除白点

 canny一下检测出来轮廓

显示全部直线

 直线提取，我的方式是提取最长的两段直线。

 在上述操作完成之后，得到了物体的粗定位直线。

但是上面的算法还是存在相应的问题，换了一个别的图像可能就检测的不准。发现问题就是出在了二值化的过程。

为了修正上方的算法的失败，使用提取外部轮廓的方式进行求取，将代码改了改。
 

#include 
#include 
#include 
#include 
#include 
#include 
using namespace cv;
using namespace std;void getPrewitt_oper(cv::Mat& getPrewitt_horizontal, cv::Mat& getPrewitt_vertical, cv::Mat& getPrewitt_Diagonal1, cv::Mat& getPrewitt_Diagonal2) {//水平方向getPrewitt_horizontal = (cv::Mat_(3, 3) << -1, -1, -1, 0, 0, 0, 1, 1, 1);//垂直方向getPrewitt_vertical = (cv::Mat_(3, 3) << -1, 0, 1, -1, 0, 1, -1, 0, 1);//对角135°getPrewitt_Diagonal1 = (cv::Mat_(3, 3) << 0, 1, 1, -1, 0, 1, -1, -1, 0);//对角45°getPrewitt_Diagonal2 = (cv::Mat_(3, 3) << -1, -1, 0, -1, 0, 1, 0, 1, 1);//逆时针反转180°得到卷积核cv::flip(getPrewitt_horizontal, getPrewitt_horizontal, -1);cv::flip(getPrewitt_vertical, getPrewitt_vertical, -1);cv::flip(getPrewitt_Diagonal1, getPrewitt_Diagonal1, -1);cv::flip(getPrewitt_Diagonal2, getPrewitt_Diagonal2, -1);
}void edge_Prewitt(cv::Mat& src, cv::Mat& dst1, cv::Mat& dst2, cv::Mat& dst3, cv::Mat& dst4, cv::Mat& dst, int ddepth, double delta = 0, int borderType = cv::BORDER_DEFAULT) {//获取Prewitt算子cv::Mat getPrewitt_horizontal;cv::Mat getPrewitt_vertical;cv::Mat getPrewitt_Diagonal1;cv::Mat getPrewitt_Diagonal2;getPrewitt_oper(getPrewitt_horizontal, getPrewitt_vertical, getPrewitt_Diagonal1, getPrewitt_Diagonal2);//卷积得到水平方向边缘cv::filter2D(src, dst1, ddepth, getPrewitt_horizontal, cv::Point(-1, -1), delta, borderType);//卷积得到4垂直方向边缘cv::filter2D(src, dst2, ddepth, getPrewitt_vertical, cv::Point(-1, -1), delta, borderType);//卷积得到45°方向边缘cv::filter2D(src, dst3, ddepth, getPrewitt_Diagonal1, cv::Point(-1, -1), delta, borderType);//卷积得到135°方向边缘cv::filter2D(src, dst4, ddepth, getPrewitt_Diagonal2, cv::Point(-1, -1), delta, borderType);//边缘强度（近似）cv::convertScaleAbs(dst1, dst1); //求绝对值并转为无符号8位图cv::convertScaleAbs(dst2, dst2);cv::convertScaleAbs(dst3, dst3); //求绝对值并转为无符号8位图cv::convertScaleAbs(dst4, dst4);dst = dst1 + dst2;}//数组从大到小排序
void reserve(int x[], int n) {int i, j, temp;for (i = 0; i < n - 1; i++) {     //一共n个元素，则需要比较n-1次for (j = 0; j < n - 1 - i; j++) {     //每一个元素需要比较的次数if (x[i] < x[i + j + 1]) {temp = x[i];x[i] = x[i + j + 1];x[i + j + 1] = temp;}}}
}int main()
{cv::Mat src = cv::imread("楼.jpg");if (src.empty()) {return -1;}cout << "??" << endl;if (src.channels() > 1) cv::cvtColor(src, src, CV_RGB2GRAY);cv::Mat dst, dst1, dst2, dst3, dst4, dst5;Mat src1 = cv::imread("楼.jpg");Mat src2 = cv::imread("楼.jpg");//medianBlur(src, src, 5);  //均值滤波GaussianBlur(src, src, Size(5, 5), 0); //高斯滤波cout << "??" << endl;//注意：要采用CV_32F，因为有些地方卷积后为负数，若用8位无符号，则会导致这些地方为0edge_Prewitt(src, dst1, dst2, dst3, dst4, dst, CV_32F);cv::namedWindow("垂直边缘", CV_WINDOW_NORMAL);imshow("垂直边缘", dst2);cout << "??" << endl;/*Mat shdjk;cv::threshold(dst2, shdjk, 25, 255, cv::THRESH_BINARY);cv::namedWindow("二值化1212", CV_WINDOW_NORMAL);imshow("二值化1212", shdjk);*///获取结构cv::Mat element1 = cv::getStructuringElement(cv::MORPH_RECT, cv::Size(3, 3));Mat out1;//进行形态学开运算操作  morphologyEx(dst2, out1, MORPH_OPEN, element1);//形态学开运算cv::namedWindow("xingtai", CV_WINDOW_NORMAL);imshow("xingtai", out1);Mat out2;//第二次进行形态学操作edge_Prewitt(out1, dst1, out1, dst3, dst4, dst, CV_32F);cv::namedWindow("垂直边缘1", CV_WINDOW_NORMAL);imshow("垂直边缘1", out1);cout << "??" << endl;/*morphologyEx(out1, out1, MORPH_OPEN, element1);//形态学开运算cv::namedWindow("xingtai1", CV_WINDOW_NORMAL);imshow("xingtai1", out1);//获取结构cv::Mat element2 = cv::getStructuringElement(cv::MORPH_ELLIPSE, cv::Size(10, 10));//进行形态学闭运算操作  morphologyEx(out1, out2, MORPH_CLOSE, element2);//形态学闭合运算cv::namedWindow("xingtai2", CV_WINDOW_NORMAL);imshow("xingtai2", out2);imwrite("xingtai2.jpg", out2);waitKey(0);*/std::vector> contours;std::vector hierarchy;findContours(out1, contours, CV_RETR_EXTERNAL, CV_CHAIN_APPROX_NONE);double maxArea = 0;int index = 0;vector maxContour;for (size_t i = 0; i < contours.size(); i++){double area = cv::contourArea(contours[i]);if (area > maxArea){maxArea = area;maxContour = contours[i];index = i;}}drawContours(src1, contours, index, Scalar(255));    // 参数cv::namedWindow("test", CV_WINDOW_NORMAL);imshow("test", src1);waitKey(0);/*Mat shdjk;cv::threshold(out1, shdjk, 10, 255, cv::THRESH_BINARY);cv::namedWindow("二值化1212", CV_WINDOW_NORMAL);imshow("二值化1212", shdjk);std::vector> contours;std::vector hierarchy;cv::findContours(shdjk, contours, hierarchy, cv::RETR_EXTERNAL, cv::CHAIN_APPROX_NONE);  //只找最外层轮廓std::vector> approxCurves(contours.size());for (int i = 0; i < contours.size(); ++i) {  //绘制逼近后的轮廓double epsilon = 0.1 * cv::arcLength(contours[i], true);cv::approxPolyDP(contours[i], approxCurves[i], epsilon, true);cv::drawContours(src1, approxCurves, i, cv::Scalar(0, 255, 0), 2);}cv::namedWindow("success", CV_WINDOW_NORMAL);imshow("success", src1);cv::waitKey();*////*Mat dhfjua;cv::threshold(out2, dhfjua, 15, 255, cv::THRESH_BINARY);cv::namedWindow("二值化000", CV_WINDOW_NORMAL);imshow("二值化000", dhfjua);*//*//膨胀运算，将细小缝隙填补上,非必要Mat kernel = getStructuringElement(0, Size(3, 3));dilate(out2, dst2, kernel);cv::namedWindow("膨胀", CV_WINDOW_NORMAL);imshow("膨胀", dst2);*//*0cv::threshold(out2, dst2, 5, 255, cv::THRESH_BINARY);cv::namedWindow("二值化", CV_WINDOW_NORMAL);imshow("二值化", dst2);cv::threshold(dst2, dst2, 5, 255, cv::THRESH_BINARY_INV);cv::namedWindow("反二值化", CV_WINDOW_NORMAL);imshow("反二值化", dst2);*//*Mat out3;//进行形态学闭运算操作  morphologyEx(dst2, out3, MORPH_CLOSE, element2);//形态学开运算cv::namedWindow("xingtai3", CV_WINDOW_NORMAL);imshow("xingtai3", out3);imwrite("xingtai3.jpg", out3);*/waitKey(0);vector lines;HoughLinesP(src1, lines, 1, CV_PI / 180, 100, 400, 30);int Length[1000] = { 0 };//存放直线长度for (size_t i = 0; i < lines.size(); i++){Vec4i I = lines[i];double x1 = I[0];double y1 = I[1];double x2 = I[2];double y2 = I[3];//筛选满足条件的点if (abs(x1 - x2) + abs(y1 - y2) > 50){Length[i] = sqrt((x1 - x2)*(x1 - x2) + (y1 - y2) * (y1 - y2));//将满足条件的点画出line(src1, Point2d(x1, y1), Point2d(x2, y2), Scalar(0, 255, 255), 2);cout << " " << "(" << x1 << "," << y1 << ")" << " " << "(" << x2 << "," << y2 << ")" << endl;//line(canny, Point2d(x1, y1), Point2d(x2, y2), Scalar(0, 255, 255), 2);}}Mat imgShow;imgShow = src1;resize(imgShow, imgShow, Size(imgShow.cols / 4, imgShow.rows / 4));imshow("imgShow", imgShow);imwrite("shuchu.png", src1);reserve(Length, 1000);for (int i = 0; i < 1000; i++) {cout << "长度" << Length[i] << endl;   //输出排序后的数组元素}for (size_t i = 0; i < lines.size(); i++){Vec4i I = lines[i];double x1 = I[0];double y1 = I[1];double x2 = I[2];double y2 = I[3];cout << "sdjk" << endl;cout << int(sqrt((x1 - x2)*(x1 - x2) + (y1 - y2) * (y1 - y2))) << endl;//筛选满足条件的点if ((int(sqrt((x1 - x2)*(x1 - x2) + (y1 - y2) * (y1 - y2))) == Length[0]) || (int(sqrt((x1 - x2)*(x1 - x2) + (y1 - y2) * (y1 - y2))) == Length[1])){//将满足条件的点画出line(src2, Point2d(x1, y1), Point2d(x2, y2), Scalar(0, 255, 255), 2);cout << "djfkljsa " << "(" << x1 << "," << y1 << ")" << " " << "(" << x2 << "," << y2 << ")" << endl;//line(canny, Point2d(x1, y1), Point2d(x2, y2), Scalar(0, 255, 255), 2);}}imgShow = src2;resize(imgShow, imgShow, Size(imgShow.cols / 4, imgShow.rows / 4));imshow("imgShow2", imgShow);imwrite("shuchu2.png", src2);waitKey(0);return 0;
}

效果还是不好，问题就是出在了相应的一个二值化的过程，因此，想到使用区域增长算法改进

#include 
#include 
#include 
#include 
#include 

终于知道是啥原因了，我在进行Prewitt算子对边缘进行粗定位检测过后，没有进行去噪处理，一定要把图像转换为二值图像，就方便多了。并且还要记住，霍夫检测的直线像素是255白线才可以，经过长时间的试错终于解决了。输入原图像如下所示，我这里使用的去噪对比了四种，但是下面这种是最好的。Opencv 非局部降噪_51CTO博客_opencv降噪Opencv 非局部降噪，opencv自带的非局部降噪算法：CV_EXPORTS_WvoidfastNlMeansDenoising(InputArraysrc,OutputArraydst,floath=3,inttemplateWindowSize=7,intsearchWindowSize=21);h是过滤强度，templateWindowSize是分块大小，searchWindowSize是搜索区域大小。应用实例intmain(){MatI..https://blog.51cto.com/u_15458280/4843576

#include
#include
using namespace std;
using namespace cv;//数组从大到小排序
void reserve(int x[], int n) {int i, j, temp;for (i = 0; i < n - 1; i++) {     //一共n个元素，则需要比较n-1次for (j = 0; j < n - 1 - i; j++) {     //每一个元素需要比较的次数if (x[i] < x[i + j + 1]) {temp = x[i];x[i] = x[i + j + 1];x[i + j + 1] = temp;}}}
}void add_salt_pepper_noise(Mat &image) {RNG rng(12345);int h = image.rows;int w = image.cols;int nums = 10000;for (int i = 0; i < nums; i++) {int x = rng.uniform(0, w);int y = rng.uniform(0, h);if (i % 2 == 1) {image.at(y, x) = Vec3b(255, 255, 255);}else {image.at(y, x) = Vec3b(0, 0, 0);}}imshow("salt pepper", image);
}void gaussian_noise(Mat &image) {Mat noise = Mat::zeros(image.size(), image.type());randn(noise, (15, 15, 15), (30, 30, 30));Mat dst;add(image, noise, dst);imshow("gaussian noise", dst);dst.copyTo(image);
}Mat convertTo3Channels(const Mat& binImg)
{Mat three_channel = Mat::zeros(binImg.rows, binImg.cols, CV_8UC3);vector channels;for (int i = 0; i < 3; i++){channels.push_back(binImg);}merge(channels, three_channel);return three_channel;
}int main(int argc, char*argv[])
{//加载图像Mat img, gray_image, dst;img = imread("垂直边缘.jpg");Mat img1 = imread("垂直边缘.jpg");//判断图像是否导入成功if (img.empty()){cout << "加载失败" << endl;return -1;}//显示图像namedWindow("original image", WINDOW_AUTOSIZE);imshow("original image", img);//转换灰度图像cvtColor(img, gray_image, COLOR_BGR2GRAY);//获取灰度图像宽度和高度int width = gray_image.cols;int height = gray_image.rows;//遍历像素值(单通道)for (int row = 0; row < height; row++){for (int col = 0; col < width; col++){int gray = gray_image.at(row, col);gray_image.at(row, col) = 255 - gray;	//图像取反};};namedWindow("inv_gray_image", WINDOW_AUTOSIZE);imshow("inv_gray_image", gray_image);Mat sh;fastNlMeansDenoising(gray_image, sh, 21, 7, 21);namedWindow("inv_gray_image1", WINDOW_AUTOSIZE);imshow("inv_gray_image1", sh);waitKey(50);//Mat s;//获取灰度图像宽度和高度width = sh.cols;height = sh.rows;//遍历像素值(单通道)for (int row = 0; row < height; row++){for (int col = 0; col < width; col++){int gray = sh.at(row, col);sh.at(row, col) = 255 - gray;	//图像取反};};namedWindow("inv_gray_image2", WINDOW_AUTOSIZE);imshow("inv_gray_image2", sh);cv::threshold(sh, sh, 50, 255, cv::THRESH_BINARY);cv::namedWindow("二值化", CV_WINDOW_NORMAL);imshow("二值化", sh);vector lines;HoughLinesP(sh, lines, 1, CV_PI / 180, 50,100, 5);int Length[100] = { 0 };//存放直线长度for (size_t i = 0; i < lines.size(); i++){Vec4i I = lines[i];double x1 = I[0];double y1 = I[1];double x2 = I[2];double y2 = I[3];//筛选满足条件的点if (abs(x1 - x2) + abs(y1 - y2) > 50){Length[i] = sqrt((x1 - x2)*(x1 - x2) + (y1 - y2) * (y1 - y2));//将满足条件的点画出line(img, Point2d(x1, y1), Point2d(x2, y2), Scalar(0, 255, 255), 2);cout << " " << "(" << x1 << "," << y1 << ")" << " " << "(" << x2 << "," << y2 << ")" << endl;//line(canny, Point2d(x1, y1), Point2d(x2, y2), Scalar(0, 255, 255), 2);}}Mat imgShow;imgShow = img;resize(imgShow, imgShow, Size(imgShow.cols / 4, imgShow.rows / 4));imshow("imgShow", imgShow);imwrite("shuchu.png", imgShow);reserve(Length, 100);for (int i = 0; i < 100; i++) {cout << "长度" << Length[i] << endl;   //输出排序后的数组元素}for (size_t i = 0; i < lines.size(); i++){Vec4i I = lines[i];double x1 = I[0];double y1 = I[1];double x2 = I[2];double y2 = I[3];cout << "sdjk" << endl;cout << int(sqrt((x1 - x2)*(x1 - x2) + (y1 - y2) * (y1 - y2))) << endl;//筛选满足条件的点if ((int(sqrt((x1 - x2)*(x1 - x2) + (y1 - y2) * (y1 - y2))) == Length[0]) || (int(sqrt((x1 - x2)*(x1 - x2) + (y1 - y2) * (y1 - y2))) == Length[1])){//将满足条件的点画出line(img1, Point2d(x1, y1), Point2d(x2, y2), Scalar(0, 255, 255), 2);cout << "djfkljsa " << "(" << x1 << "," << y1 << ")" << " " << "(" << x2 << "," << y2 << ")" << endl;//line(canny, Point2d(x1, y1), Point2d(x2, y2), Scalar(0, 255, 255), 2);}}imgShow = img1;resize(imgShow, imgShow, Size(imgShow.cols / 4, imgShow.rows / 4));imshow("imgShow2", imgShow);imwrite("shuchu2.png", imgShow);waitKey(0);return 0;
};

结果图如下所示：

 ​​​​​​​

 终于弄出来了，去干饭。