C#：用OpenCV实现缺陷检测
2023-02-17 本文已影响0人
大龙10
一、简介
- 机器视觉应用场景中缺陷检测的应用是非常广泛的，通常涉及各个行业、各种缺陷类型。
- 纺织物的缺陷检测，缺陷类型包含脏污、油渍、线条破损三种，这三种缺陷与LCD屏幕检测的缺陷很相似，处理方法也可借鉴。
- 使用OpenCV中的FindContours函数可以 实现纺织物缺陷检测(脏污、油渍、线条破损缺陷)。
二、FindContours函数
- FindContours函数
找轮廓
void findContours( InputOutputArray image,
OutputArrayOfArrays contours,
OutputArray hierarchy, int mode,
int method, Point offset = Point());
-
InputOutputArray image：输入图像是8位单通道的图像（256级灰度图）。
其中像素点的非0灰度值被当成1(转化后即为255)，0值保持0，所以输入图像被当成一个二值图像对待。
可以用compare() , inRange() , threshold() , adaptiveThreshold() , Canny()或者其他方法来从灰度图或者彩色图中生成二值图像。该函数在提取轮廓的过程中会改变图像。
如果第4个参数 mode 为 CV_RETR_CCOMP 或者
CV_RETR_FLOODFILL，输入图像也可以是32位的整型图像(CV_32SC1)。 -
OutputArrayOfArrays contours: 检测到的轮廓
Each contour is stored as a vector of points. 每个轮廓会被存储为vector<Point>
所以 contours 的类型是vector<vector<Point>>。 -
OutputArray hierarchy: 可选的输出向量，包含图像的拓扑信息
It has as many elements as the number of contours. 元素个数 = 轮廓数
对于第 i 个轮廓contours[i]，hierarchy 的以下元素分别表示
hierarchy[i][0]: the next contour at the same hierarchical level
hierarchy[i][1]: the previous contour at the same hierarchical level
hierarchy[i][2]: the first child contour
hierarchy[i][3]: the parent contour
hierarchy 的这些元素的原始值为0，如果不存在，置为负数 -
int mode: Contour retrieval mode 取回轮廓模式（复杂度依次增加）
三、检测条件
先进行二值化、高斯滤波、平滑等处理。再进行轮廓分析。
- 脏污
轮廓圆弧长度大于1 - 油渍
轮廓面积大于50 -
线条破损
轮廓圆弧长度大于10
脏污检测
四、程序源码
using System;
using System.Collections.Generic;
using System.ComponentModel;
using System.Data;
using System.Drawing;
using System.Linq;
using System.Runtime.CompilerServices;
using System.Runtime.InteropServices;
using System.Text;
using System.Threading.Tasks;
using System.Windows.Forms;
using Sunny.UI;
using OpenCvSharp;
using OpenCvSharp.Extensions;
namespace Ky_FindContours
{
public partial class Form1 : UIForm
{
public Form1()
{
InitializeComponent();
this.SetStyle(ControlStyles.OptimizedDoubleBuffer | ControlStyles.ResizeRedraw | ControlStyles.AllPaintingInWmPaint, true);
}
private Image image = null;
private Mat dst = new Mat();
private Mat src_img;
string filePath = "";
private List<Mat> reList = new List<Mat>();
private int step = 1;
private void openImage_Click(object sender, EventArgs e)
{
OpenFileDialog openFileDialog = new OpenFileDialog();
openFileDialog.Title = "选择操作的图片";
openFileDialog.Filter = "图片 *.jpg|*.jpg|图像*.png|*.png";
if (openFileDialog.ShowDialog() == DialogResult.OK)
{
filePath = openFileDialog.FileName;
image = Image.FromFile(filePath);
src_img = Cv2.ImRead(filePath);
Mat tem1 = new Mat();
src_img.CopyTo(tem1);
if (reList.Count > 0)
{
reList[0] = tem1;
}
else
{
reList.Add(tem1);
}
}
if (filePath != "")
{
picBoxShowDel.Image = image;
picShowOri.Image = image;
}
}
/// <summary>
/// 脏污缺陷检测
/// </summary>
/// <param name="img">测试图像</param>
/// <returns>结果图</returns> //也可设置bool类型表示OK或NG
static Mat DirtyDetection(Mat img)
{
Mat result = img.Clone();
Mat gray = new Mat();
Cv2.CvtColor(img, gray, ColorConversionCodes.BGR2GRAY);
Cv2.GaussianBlur(gray, gray, new OpenCvSharp.Size(7, 7), 0);
Cv2.Canny(gray, gray, 10, 30);
OpenCvSharp.Point[][] contours; //轮廓查找结果变量
HierarchyIndex[] hierarchy; //轮廓拓扑结构变量
Cv2.FindContours(gray, out contours, out hierarchy, RetrievalModes.External,
ContourApproximationModes.ApproxNone);
//Console.WriteLine("contour_size = {0}", contours.Length); //输出轮廓个数
for (int i = 0; i < contours.Length; i++)
{
double length = Cv2.ArcLength(contours[i], true);
if (length >= 1)
Cv2.DrawContours(result, contours, i, new Scalar(0, 0, 255), 2);
}
return result;
}
/// <summary>
/// 油污缺陷检测
/// </summary>
/// <param name="img">测试图像</param>
/// <returns>结果图</returns> //也可设置bool类型表示OK或NG
static Mat OilDetection(Mat img)
{
Mat result = img.Clone();
Mat imgLab = new Mat();
Cv2.CvtColor(img, imgLab, ColorConversionCodes.BGR2Lab);
Mat[] labArray = Cv2.Split(imgLab); //L, a, b
Mat blur = new Mat();
Mat thres = new Mat();
Cv2.GaussianBlur(labArray[2], blur, new OpenCvSharp.Size(3, 3), 0); //b通道
Cv2.Threshold(blur, thres, 130, 255, ThresholdTypes.Binary);
Mat element = Cv2.GetStructuringElement(MorphShapes.Rect, new OpenCvSharp.Size(3, 3), new OpenCvSharp.Point(-1, -1));
Cv2.MorphologyEx(thres, thres, MorphTypes.Open, element, new OpenCvSharp.Point(-1, -1), 1,
BorderTypes.Default, new Scalar());
OpenCvSharp.Point[][] contours; //轮廓查找结果变量
HierarchyIndex[] hierarchy; //轮廓拓扑结构变量
Cv2.FindContours(thres, out contours, out hierarchy, RetrievalModes.External,
ContourApproximationModes.ApproxNone);
//Console.WriteLine("contour_size = {0}", contours.Length); //输出轮廓个数
for (int i = 0; i < contours.Length; i++)
{
double area = Cv2.ContourArea(contours[i]);
if (area >= 50)
Cv2.DrawContours(result, contours, i, new Scalar(0, 0, 255), 2);
}
return result;
}
/// <summary>
/// 线条破损缺陷检测
/// </summary>
/// <param name="img">测试图像</param>
/// <returns>结果图</returns> //也可设置bool类型表示OK或NG
static Mat LineDefectDetection(Mat img)
{
Mat result = img.Clone();
Mat imgLab = new Mat();
Cv2.CvtColor(img, imgLab, ColorConversionCodes.BGR2Lab);
Mat[] labArray = Cv2.Split(imgLab); //L, a, b
Mat blur = new Mat();
Mat edged = new Mat();
Cv2.GaussianBlur(labArray[2], blur, new OpenCvSharp.Size(3, 3), 0); //b通道
Cv2.Canny(blur, edged, 5, 10);
OpenCvSharp.Point[][] contours; //轮廓查找结果变量
HierarchyIndex[] hierarchy; //轮廓拓扑结构变量
Cv2.FindContours(edged, out contours, out hierarchy, RetrievalModes.External,
ContourApproximationModes.ApproxNone);
//Console.WriteLine("contour_size = {0}", contours.Length); //输出轮廓个数
for (int i = 0; i < contours.Length; i++)
{
double length = Cv2.ArcLength(contours[i], true);
if (length >= 10)
Cv2.DrawContours(result, contours, i, new Scalar(0, 0, 255), 2);
}
return result;
}
private void uiButton1_Click(object sender, EventArgs e)
{
Mat zw_result = DirtyDetection(src_img); //脏污缺陷检测
picBoxShowDel.Image = zw_result.ToBitmap();
}
private void uiButton2_Click(object sender, EventArgs e)
{
Mat yw_result = OilDetection(src_img); //油污缺陷检测
picBoxShowDel.Image = yw_result.ToBitmap();
}
private void uiButton3_Click(object sender, EventArgs e)
{
Mat yw_result = LineDefectDetection(src_img); //线条破损缺陷检测
picBoxShowDel.Image = yw_result.ToBitmap();
}
}
}
油污检测
五、参考资料
博客：https://blog.51cto.com/stq054188/5543992
来源公众号：OpenCV与AI深度学习
