如何检测两行或图像指定区域之间的轮廓数

Question

我正在使用 OpenCV C++ 来查找视频中的轮廓。我想计算视频中指定区域或视频中绘制的两条线之间存在的轮廓的数量。例如，轮廓流在视频中移动，我想在它们到达视频中的特定区域时对它们进行计数。当他们离开视频中的特定区域时，我会减少计数。我知道一些基本的东西来找到轮廓，计算面积等。但我没有得到任何编程技巧来计算指定区域内的轮廓数。请帮助我了解相关主题和一些编程技巧。 （我不想使用 cvBlob.h 库）

基本上我是在计算进入该地区的汽车数量。如果输入汽车，我将增加计数，如果它离开该区域，则我将减少计数。

Answer 1

您可以将找到的轮廓近似为多边形或圆形：

for( int i = 0; i < contours.size(); i++ )
 { approxPolyDP( Mat(contours[i]), contours_poly[i], 3, true );
   boundRect[i] = boundingRect( Mat(contours_poly[i]) )
 }

然后使用直线方程y=a，比较矩形的角坐标或圆的center+radius，以确定轮廓是否通过直线。

Answer 2

1.使用您的 Mat 图像的一部分。（如果您的 ROI 是矩形）

Mat src;  // suppose this is your source frame
Mat src_of_interest = src(Rect(Point(x1, y1), Point(x2, y2)));

// Do the rest of finding contour..

vector<vector<Point>> contours;
vector<Vec4i> hierarchy;

findContours(src_of_interest.clone(), contours, hierarchy, CV_RETR_TREE, CV_CHAIN_APPROX_SIMPLE);

//now you can do your job with contour vector.    
int count = contours.size();

Mat dst = Mat::zeros(src.size(), CV_8UC3);
for(int i=0; i< count; i++)
{
    drawContours(dst, contours, i, CV_RGB(255, 0, 0));   // draw contour in red
}

2. 如果您的感兴趣区域不是矩形，请尝试以下方法：

vector<Point> contour_of_interest;   // this contour is where you want to check
vector<vector<Point>> contours;   // this is the contours you found

Mat dst = Mat::zeros(src.size(), CV_8U);
for(int i=0; i< count; i++)
{
    drawContours(dst, contours, i, Scalar(255));   // set contour area to 255
}

Mat roi = Mat::zeros(src.size(), CV_8U);
vector<vector<Point>> coi_vector;
coi_vector.push_back(contour_of_interest);
drawContours(roi, coi_vector, 0, Scalar(255));

Mat and = dst & roi;    // where 2 mats are both TRUE(not zero)

findContours(and.clone(), contours, hierarchy, CV_RETR_TREE, CV_CHAIN_APPROX_SIMPLE);

//now you can do your job with contour vector.    
int count = contours.size();

3.如果你想计算4个点之间的轮廓，试试这个方法：

vector<Point> contour_of_interest;   // this contour is the area where you want to check
Point p1(x1, y1);
Point p2(x1, y2);
Point p3(x2, y2);
Point p4(x2, y1);
contour_of_interest.push_back(p1);
contour_of_interest.push_back(p2);
contour_of_interest.push_back(p3);
contour_of_interest.push_back(p4);

vector<vector<Point>> coi_list;
coi_list.push_back(contour_of_interest);

Mat mask = Mat:zeros(src.size(), CV_8U);
drawContours(mask, coi_list, 0, Scalar(255));

Mat src;  // suppose this is your source frame
Mat src_of_interest = src & mask;  // remove any pixels outside mask area

// Do the rest of finding contour..

vector<vector<Point>> contours;
vector<Vec4i> hierarchy;

findContours(src_of_interest.clone(), contours, hierarchy, CV_RETR_TREE, CV_CHAIN_APPROX_SIMPLE);

//now you can do your job with contour vector.    
int count = contours.size();

Mat dst = Mat::zeros(src.size(), CV_8UC3);
for(int i=0; i< count; i++)
{
    drawContours(dst, contours, i, CV_RGB(255, 0, 0));   // draw contour in red
}