前些天發現了一個巨牛的人工智能學習網站,通俗易懂,風趣幽默,忍不住分享一下給大家:
點選跳轉
目錄
一,輪廓檢測
二,邊緣檢測
一,輪廓檢測
以下圖為例:
![](https://img.laitimes.com/img/_0nNw4CM6IyYiwiM6ICdiwiI0gTMx81dsQWZ4lmZf1GLlpXazVmcvwFciV2dsQXYtJ3bm9CX9s2RkBnVHFmb1clWvB3MaVnRtp1XlBXe0xCMy81dvRWYoNHLwEzX5xCMx8FesU2cfdGLwMzX0xiRGZkRGZ0Xy9GbvNGLpZTY1EmMZVDUSFTU4VFRR9Fd4VGdsYTMfVmepNHLrJXYtJXZ0F2dvwVZnFWbp1zczV2YvJHctM3cv1Ce-cmbw5yNxMjM3MTO0EjMjJjZklDNyYzX4QzM1EDM0AzLcRDMyIDMy8CXn9Gbi9CXzV2Zh1WavwVbvNmLvR3YxUjLyM3Lc9CX6MHc0RHaiojIsJye.png)
灰階圖直接提取輪廓:
int main()
{
Mat img = imread("D:/1.png", 0);
resize(img, img, Size(0, 0), 0.5, 0.5);
Mat src=img;
cv::imshow("src", src);
std::vector<std::vector<Point>> contours;
std::vector<Vec4i> hierarchy;
findContours(src, contours, hierarchy, RETR_LIST, CHAIN_APPROX_NONE, Point(0, 0));
cout << contours.size()<<endl;
//sort(contours.begin(), contours.end(), cmp< Point>);
//for (int i = 0; i < contours.size(); i++)cout << contours[i].size() << " ";
src = 0;
cv::drawContours(src, contours, -1, cv::Scalar::all(255));
cv::imshow("Contours", src);
cv::waitKey(0);
return 0;
}
得到的輪廓數量是1,即整個圖的最外面的輪廓。
優化思路:先進行一定的二值化處理。
int main()
{
Mat img = imread("D:/1.png", 0);
resize(img, img, Size(0, 0), 0.5, 0.5);
Mat src;
threshold(img, src, 200, 255, THRESH_TRUNC);
threshold(src, src, 100, 255, THRESH_TOZERO);
cv::imshow("src", src);
std::vector<std::vector<Point>> contours;
std::vector<Vec4i> hierarchy;
findContours(src, contours, hierarchy, RETR_LIST, CHAIN_APPROX_NONE, Point(0, 0));
cout << contours.size()<<endl;
//sort(contours.begin(), contours.end(), cmp< Point>);
//for (int i = 0; i < contours.size(); i++)cout << contours[i].size() << " ";
src = 0;
cv::drawContours(src, contours, -1, cv::Scalar::all(255));
cv::imshow("Contours", src);
cv::waitKey(0);
return 0;
}
一般提取輪廓前,先做一下邊緣檢測更好。
邊緣檢測偏向于圖像中像素點的變化,輪廓檢測更偏向于關注上層語義對象。
二,邊緣檢測
利用Canny算子做邊緣檢測:
int main()
{
Mat img = imread("D:/1.png", 0);
resize(img, img, Size(0, 0), 0.5, 0.5);
Mat src;
threshold(img, src, 200, 255, THRESH_TRUNC);
threshold(src, src, 100, 255, THRESH_TOZERO);
Canny(src, src, 200, 100, 3);
cv::imshow("src", src);
cv::waitKey(0);
return 0;
}
邊緣檢測結果很清晰。