Halcon实例转OpenCVSharp--粘连物体分割与计数(基于形态学+连通域计算) (附源码)

本文作者Color Space，文章未经作者允许禁止转载！

本文将介绍OpenCVSharp应用实例--粘连物体分割与计数(基于形态学+连通域计算)！

本文使用的OpenCVSharp版本---》v4.5.3。

实现目标①：将下图中的糖豆分割出来并计算数量。

预期效果图：

实现目标②：将下图中的工件分割出来并计算数量。

预期效果图： 

实现步骤：

【1】Otsu阈值或区间阈值做二值化处理；

//Otsu二值化
Cv2.Threshold(grayImg, thresImg, 0, 255, ThresholdTypes.Otsu);

//或者区间二值化
//Cv2.Threshold(grayImg, thresImg, 106, 255, ThresholdTypes.Binary);

【2】创建圆形结构元素 + 腐蚀操作 --》将粘连区域断开

//创建圆形结构元素
Mat kernal = Mat.Zeros(new Size(15, 15), MatType.CV_8U);
Cv2.Circle(kernal, new Point(7, 7), 7, new Scalar(1, 1, 1), -1);
Cv2.Erode(thresImg, thresImg, kernal);

【3】计算连通域 + 连通域膨胀操作

//获得连通域及其属性
ConnectedComponents cc = Cv2.ConnectedComponentsEx(thresImg, PixelConnectivity.Connectivity8, ConnectedComponentsAlgorithmsTypes.Default);
int count = cc.Blobs.Count; //连通域数量0~N, 0对应背景，前景连通域数量N-1
cc.RenderBlobs(imgCopy); //绘制连通域到指定图像
Cv2.ImShow("RenderBlobs", imgCopy);
Cv2.Dilate(imgCopy, imgCopy, kernal);
Cv2.ImShow("dilate", imgCopy);

【4】连通域膨胀结果和原图按权重叠加 + 标注连通域中心 + 输出连通域数量

Mat result = new Mat();
Cv2.AddWeighted(srcImg, 0.8, imgCopy, 0.5, 0, result);
for (int i = 1; i < count; i++) //Blobs[0]对应背景, 这里跳过
{
    Point center = new Point((int)cc.Blobs[i].Centroid.X, (int)cc.Blobs[i].Centroid.Y);
    Cv2.DrawMarker(result, center, new Scalar(0, 0, 255), MarkerTypes.TiltedCross,
                                               10, 2);
}
Cv2.PutText(result, "count="+(count-1).ToString(), new Point(260, 20),         
            HersheyFonts.HersheySimplex, 0.7, new Scalar(0, 255, 0), 2);
Cv2.ImShow("result", result);
Cv2.WaitKey(0);

OpenCVSharp完整源码：

//公众号：OpenCV与AI深度学习
//作者：Color Space
//CSDN: https://blog.csdn.net/stq054188?spm=1010.2135.3001.5343
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
using OpenCvSharp;

namespace Segment_And_Count_Objects
{
    class Program
    {
        static void Main(string[] args)
        {
            Mat srcImg = Cv2.ImRead("pellets.png");
            Mat imgCopy = srcImg.Clone();
            Cv2.ImShow("src", srcImg);
            Mat grayImg = new Mat();
            Mat thresImg = new Mat();
            Cv2.CvtColor(srcImg, grayImg, ColorConversionCodes.BGR2GRAY);
            Cv2.Threshold(grayImg, thresImg, 0, 255, ThresholdTypes.Otsu);
            //Cv2.Threshold(grayImg, thresImg, 106, 255, ThresholdTypes.Binary);
            Cv2.ImShow("threshold", thresImg);
            //创建圆形结构元素
            Mat kernal = Mat.Zeros(new Size(15, 15), MatType.CV_8U);
            Cv2.Circle(kernal, new Point(7, 7), 7, new Scalar(1, 1, 1), -1);
            Cv2.Erode(thresImg, thresImg, kernal);
            Cv2.ImShow("erode", thresImg);
            //获得连通域及其属性
            ConnectedComponents cc = Cv2.ConnectedComponentsEx(thresImg, PixelConnectivity.Connectivity8, ConnectedComponentsAlgorithmsTypes.Default);
            int count = cc.Blobs.Count; //连通域数量0~N, 0对应背景，前景连通域数量N-1
            cc.RenderBlobs(imgCopy); //绘制连通域到指定图像
            Cv2.ImShow("RenderBlobs", imgCopy);
            Cv2.Dilate(imgCopy, imgCopy, kernal);
            Cv2.ImShow("dilate", imgCopy);
            //foreach (var blob in cc.Blobs) //获取每个连通域属性
            Mat result = new Mat();
            Cv2.AddWeighted(srcImg, 0.8, imgCopy, 0.5, 0, result);
            for (int i = 1; i < count; i++) //Blobs[0]对应背景, 这里跳过
            {
                Point center = new Point((int)cc.Blobs[i].Centroid.X, (int)cc.Blobs[i].Centroid.Y);
                Cv2.DrawMarker(result, center, new Scalar(0, 0, 255), MarkerTypes.TiltedCross,
                                               10, 2);
            }
            Cv2.PutText(result, "count="+(count-1).ToString(), new Point(260, 20), HersheyFonts.HersheySimplex, 0.7, new Scalar(0, 255, 0), 2);
            Cv2.ImShow("result", result);
            Cv2.WaitKey(0);
        }
    }
}

实例二对应的代码除了部分参数不同，其他代码都不变，下面是效果图和源码，大家可以自己学习体会：

 OpenCVSharp源码：

8?spm=1010.2135.3001.5343
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
using OpenCvSharp;

namespace Segment_And_Count_Objects
{
    class Program
    {
        static void Main(string[] args)
        {
            Mat srcImg = Cv2.ImRead("B0.jpg");
            Mat imgCopy = srcImg.Clone();
            Cv2.ImShow("src", srcImg);
            Mat grayImg = new Mat();
            Mat thresImg = new Mat();
            Cv2.CvtColor(srcImg, grayImg, ColorConversionCodes.BGR2GRAY);
            Cv2.Threshold(grayImg, thresImg, 70, 255, ThresholdTypes.Binary);
            Cv2.ImShow("threshold", thresImg);
            //创建圆形结构元素
            Mat kernal = Mat.Zeros(new Size(53, 53), MatType.CV_8U);
            Cv2.Circle(kernal, new Point(26, 26), 26, new Scalar(1, 1, 1), -1);
            Cv2.Erode(thresImg, thresImg, kernal);
            Cv2.ImShow("erode", thresImg);
            //获得连通域及其属性
            ConnectedComponents cc = Cv2.ConnectedComponentsEx(thresImg, PixelConnectivity.Connectivity8, ConnectedComponentsAlgorithmsTypes.Default);
            int count = cc.Blobs.Count; //连通域数量0~N, 0对应背景，前景连通域数量N-1
            cc.RenderBlobs(imgCopy); //绘制连通域到指定图像
            Cv2.ImShow("RenderBlobs", imgCopy);
            Cv2.Dilate(imgCopy, imgCopy, kernal);
            Cv2.ImShow("dilate", imgCopy);
            //foreach (var blob in cc.Blobs) //获取每个连通域属性
            Mat result = new Mat();
            Cv2.AddWeighted(srcImg, 0.8, imgCopy, 0.5, 0, result);
            for (int i = 1; i < count; i++) //Blobs[0]对应背景, 这里跳过
            {
                Point center = new Point((int)cc.Blobs[i].Centroid.X, (int)cc.Blobs[i].Centroid.Y);
                Cv2.DrawMarker(result, center, new Scalar(0, 0, 255), MarkerTypes.TiltedCross,
                                               20, 2);
            }
            Cv2.PutText(result, "count=" + (count - 1).ToString(), new Point(20, 30), HersheyFonts.HersheySimplex, 1.0, new Scalar(0, 255, 0), 2);
            Cv2.ImShow("result", result);
            Cv2.WaitKey(0);
        }
    }
}

更多内容请关注微信公众号：OpenCV与AI深度学习 

​