本文所述方法可以检测同一图像中的多个圆形（准确的说，应该是闭合图像）。
在Matlaba中可以实现。
附录效果图：
%颗粒圆度clear;close all;%%%读取源图像I = imread('999.png');figure;imshow(I);%%%灰度化、取反h = rgb2gray(I);figure;imshow(h);%灰度图像h = imcomplement(h);%取反figure;imshow(h);%%%中值滤波、二值化h = medfilt2(h,[4,4]);bw = im2bw(h,graythresh(h));%%%消除噪点se = strel('disk',2);bw = imclose(bw,se);figure;imshow(bw);%%%填补闭合图形，填充色为白色bw = imfill(bw,'holes');%%%边界寻找[B,L] = bwboundaries(bw,'noholes');% 为每个闭合图形设置颜色显示figure;imshow(label2rgb(L, @jet, [.5 .5 .5]))hold onfor k = 1:length(B)boundary = B{k};plot(boundary(:,2), boundary(:,1), 'w', 'LineWidth', 2)end%%%计算面积stats = regionprops(L,'Area','Centroid'); threshold = 0.94; % 循环处理每个边界，length(B)是闭合图形的个数,即检测到的陶粒对象个数for k = 1:length(B) % 获取边界坐标'boundary = B{k}; % 计算周长delta_sq = diff(boundary).^2;perimeter = sum(sqrt(sum(delta_sq,2))); % 对标记为K的对象获取面积area = stats(k).Area; % 圆度计算公式4*PI*A/P^2metric = 4*pi*area/perimeter^2; % 结果显示metric_string = sprintf('%2.2f',metric); % 用一个黑色小圆圈标记圆度大于threshold = 0.94 的对象if metric > thresholdcentroid = stats(k).Centroid;plot(centroid(1),centroid(2),'ko');end%设置显示字体text(boundary(1,2)-35,boundary(1,1)+13,metric_string,'Color','y',...'FontSize',14,'FontWeight','bold'); end title(['圆度识别结果，越圆越接近1，']);
