【opnecv】检测桌子上多余的物品

世_生

于 2025-06-04 17:06:18 发布

阅读量230

点赞数 4

CC 4.0 BY-SA版权

分类专栏： opencv 文章标签：计算机视觉 opencv

本文链接：https://blog.csdn.net/qq_53505979/article/details/148429487

opencv 专栏收录该内容

2 篇文章

订阅专栏

效果展示

在这里插入图片描述

实现原理

1、转化成灰度图

 Mat image = imread("C:\\Users\\chenxinyu\\Desktop\\物品.jpg",IMREAD_COLOR);
 if (image.empty())
     return -1;
 Mat outImage;
 cvtColor(image, outImage, COLOR_BGR2GRAY);

2、绘制轮廓

在绘制轮廓之前，我们需要进行高斯模糊降噪。

 Mat outBlur,outCanny;
 GaussianBlur(outImage, outBlur,Size(3,3),1.5);
 Canny(outBlur, outCanny,100,150);

减少无关噪声对轮廓的影响。
在这里插入图片描述

3、二值化操作

对原图像进行二值化操作，再与轮廓图相加，丰富饱满轮廓图。

 Mat outThreshold;
 threshold(outImage, outThreshold, 120, 255, THRESH_BINARY_INV);
 Mat outThreshold1;
 threshold(outImage, outThreshold1, 220, 255, THRESH_BINARY);
 Mat outOr;
 bitwise_or(outCanny, outThreshold, outOr);
 bitwise_or(outOr, outThreshold1, outOr);

注：第二次二值化是为针对图像中白色蛋壳。

在这里插入图片描述

4、膨胀和腐蚀

去掉多余的线条，让图像更加饱满。

Mat outElement;
outElement=getStructuringElement(MORPH_ELLIPSE, Size(3, 3));
morphologyEx(outOr, outOr, MORPH_CLOSE, outElement);
morphologyEx(outOr, outOr, MORPH_OPEN, outElement);

在这里插入图片描述

5、找轮廓、标识

vector<vector<Point>> contours;
vector<Vec4i> hierarchy;
findContours(outOr, contours, hierarchy, RETR_EXTERNAL, CHAIN_APPROX_SIMPLE);
for (size_t i = 0; i < contours.size(); i++) {
     // 计算面积
     double area = contourArea(contours[i]); 
     // 标注瑕疵
     Rect bbox = boundingRect(contours[i]);
     rectangle(image, bbox, Scalar(0, 0, 255), 2);  // 红色边界框
}

在这里插入图片描述

整体代码

int demo4()
{
    Mat image = imread("C:\\Users\\chenxinyu\\Desktop\\物品.jpg",IMREAD_COLOR);
    if (image.empty())
        return -1;

    imshow("111.jpg", image);

    Mat outImage;
    cvtColor(image, outImage, COLOR_BGR2GRAY);
    imshow("222.jpg", outImage);

    ////画轮廓
    Mat outBlur;
    GaussianBlur(outImage, outBlur,Size(3,3),1.5);
    imshow("333.jpg", outBlur);
    Mat outCanny;
    Canny(outBlur, outCanny,100,150);
    imshow("444.jpg", outCanny);

    Mat outCanny1;
    Canny(outImage, outCanny1, 100, 150);
    imshow("text1.jpg", outCanny1);

    //二值分化
    Mat outThreshold;
    threshold(outImage, outThreshold, 120, 255, THRESH_BINARY_INV);
    imshow("555.jpg", outThreshold);

    Mat outThreshold1;
    threshold(outImage, outThreshold1, 220, 255, THRESH_BINARY);
    imshow("text.jpg", outThreshold1);


    Mat outOr;
   
    bitwise_or(outCanny, outThreshold, outOr);
    bitwise_or(outOr, outThreshold1, outOr);
    imshow("666.jpg", outOr);

    Mat outElement;
    outElement=getStructuringElement(MORPH_ELLIPSE, Size(3, 3));
    morphologyEx(outOr, outOr, MORPH_CLOSE, outElement);
    morphologyEx(outOr, outOr, MORPH_OPEN, outElement);
   
    imshow("777.jpg", outOr);

    vector<vector<Point>> contours;
    vector<Vec4i> hierarchy;
    findContours(outOr, contours, hierarchy, RETR_EXTERNAL, CHAIN_APPROX_SIMPLE);
    for (size_t i = 0; i < contours.size(); i++) {
        // 忽略小面积区域（噪声）
        double area = contourArea(contours[i]);
       
        // 标注瑕疵
        Rect bbox = boundingRect(contours[i]);
        rectangle(image, bbox, Scalar(0, 0, 255), 2);  // 红色边界框
    }
    imshow("888.jpg", image);
    waitKey(0);
}