openCV 轮廓查找-测量物体尺寸

一，利用openCV的findContours轮廓查找功能，用已知物体的尺寸（比如硬币）作为参考，根据实际尺寸与像素尺寸的比列，求出图片中物体的实际大小。存在的问题有两个：

• 图片的阴影导致轮廓不准确，除了多给点光之外没有想到怎么消除阴影。
• 拍摄的角度要求垂直物体，这个也很难做到，可以靠工装固定，目前不知道怎么用算法来规避。

二，流程

三，代码

#include "stdafx.h"
#include <iostream>

#include "opencv2/core/core.hpp"
#include "opencv2/imgproc/imgproc.hpp"
#include "opencv2/video/background_segm.hpp"
#include "opencv2/highgui/highgui.hpp"
#include <opencv\cxcore.hpp>
#include <stdio.h>
using namespace std;
using namespace cv;

Mat g_srcImage;
Mat g_grayImage;

const int g_dReferWidth=19;//mm 
double g_dPixelsPerMetric;
vector<vector<cv::Point>> g_vContours;
vector<Vec4i> g_vHierarchy;
bool g_bFirst=true;

static cv::Point2f midpoint(cv::Point2f& ptA, cv::Point2f& ptB);//求中点 
static float getDistance(Point2f pointA, Point2f pointB);//求距离
static bool ContoursSortFun(vector<cv::Point> contour1, vector<cv::Point> contour2);//按照 x坐标 排序
//-----------------------------------【main( )函数】--------------------------------------------
//      描述：控制台应用程序的入口函数，我们的程序从这里开始
//-------------------------------------------------------------------------------------------------
int main(int argc, const char** argv)
{
	system("color 1F");

	g_srcImage = imread("1.jpg", 1);
	//灰度 降低计算量
	cvtColor(g_srcImage, g_grayImage, COLOR_BGR2GRAY);
	
	//高斯滤波 降噪
	GaussianBlur(g_grayImage, g_grayImage, Size(7, 7), 0);
	imshow("高斯滤波", g_grayImage);
	
	//边缘检测
	Canny(g_grayImage, g_grayImage, 50,100);

	Mat element = getStructuringElement(MORPH_RECT, Size(15, 15)); //隔开物体
	dilate(g_grayImage, g_grayImage, element);//膨胀
	erode(g_grayImage, g_grayImage, element);//腐蚀
	imshow("形态学", g_grayImage);

	//寻找轮廓
	findContours(g_grayImage, g_vContours, g_vHierarchy, CV_RETR_EXTERNAL, CHAIN_APPROX_SIMPLE);
	
	std::sort(g_vContours.begin(),g_vContours.end(), ContoursSortFun);//按照从左到右 排序
	for (unsigned i = 0; i < g_vContours.size(); i++) {
		
		if (contourArea(g_vContours[i]) < 100)//面积太小 则忽略
			continue;

		RotatedRect box = minAreaRect(g_vContours[i]);
		Point2f boxPoints[4];
		box.points(boxPoints);

		Point2f pointA = midpoint(boxPoints[0], boxPoints[1]);
		Point2f pointB = midpoint(boxPoints[1], boxPoints[2]);
		Point2f pointC = midpoint(boxPoints[2], boxPoints[3]);
		Point2f pointD = midpoint(boxPoints[3], boxPoints[0]);

		circle(g_srcImage, pointA, 2, Scalar(0, 0, 255));
		circle(g_srcImage, pointB, 2, Scalar(0, 0, 255));
		circle(g_srcImage, pointC, 2, Scalar(0, 0, 255));
		circle(g_srcImage, pointD, 2, Scalar(0, 0, 255));

		line(g_srcImage, pointA, pointC, Scalar(255, 0, 0));
		line(g_srcImage, pointD, pointB, Scalar(255, 0, 0));

		double dWidth = getDistance(pointA, pointC);
		double dHeight = getDistance(pointD, pointB);
		if (g_bFirst) {
			g_dPixelsPerMetric = dWidth/g_dReferWidth; //计算像素与 实际大小的比列
			cout << "pixelPerMetric:" << dWidth <<" "<< g_dReferWidth <<"  "<< g_dPixelsPerMetric;
			g_bFirst = false;
		}

		cout << "dWidth" << dWidth <<"   "<< dHeight<< "      "<<dWidth / g_dPixelsPerMetric<<"    "<< dHeight / g_dPixelsPerMetric;
		putText(g_srcImage, cv::format("(%.0f,%.0f)", dWidth/g_dPixelsPerMetric,dHeight/g_dPixelsPerMetric), boxPoints[2],FONT_HERSHEY_COMPLEX,0.5,Scalar(0,0,255));

		for (int i = 0; i <=3; i++)
		{
			line(g_srcImage, boxPoints[i], boxPoints[(i + 1) % 4], Scalar(0, 255, 0));
		}
	}


	cv::namedWindow("效果", CV_WINDOW_AUTOSIZE);
	cv::imshow("效果", g_srcImage);

	waitKey(0);
	return 0;
}

Point2f midpoint(Point2f& ptA, Point2f& ptB) {
	return Point2f((ptA.x+ ptB.x)*0.5,(ptA.y + ptB.y)*0.5);
}

float getDistance(Point2f pointA, Point2f pointB)
{
	float distance;
	distance = powf((pointA.x - pointB.x), 2) + powf((pointA.y - pointB.y), 2);
	distance = sqrtf(distance);
	return distance;
}

bool ContoursSortFun(vector<cv::Point> contour1, vector<cv::Point> contour2) {
	return  (contour1[0].x<contour2[0].x); // a.x < b.x;
}

四，效果

参考：使用OpenCV测量图像中物体的大小_-出发-的博客-CSDN博客_opencv测量物体尺寸