自动对焦搜索算法(未完)

自动对焦搜索算法

先黄金搜索再爬山搜索

1、爬山搜索法

2、黄金搜索法


3、代码

// AutoFocus.cpp : 定义控制台应用程序的入口点。
//

#include "stdafx.h"
#include <iostream>
#include <vector>
#include <math.h>
#include <halconcpp.h>

using namespace HalconCpp;
using namespace std;

vector<int> GoldenSectionSearch(double m[], int a, int b, double k, int T);
int HillClimbingSearch(double m[], int c, int d);
int EvaluationImage(int k);

//自动对焦主函数
int _tmain(int argc, _TCHAR* argv[])
{
	double m[30] = { 0.5, 1.7, 3.4, 5.5, 9.7, 13.5, 17.2, 20.6, 23.1, 25.9, 32.3, 38.7, 43.3, 50.5, 44.1, 37.7, 29.3, 22.7, 19.3, 13.6, 11.4, 8.5, 6.6, 4.7, 3.1, 1.9, 0.9, 0.7 ,0.6, 0.4 };
	int a = 0, b = 29;
	vector<int> tmp;
	tmp = GoldenSectionSearch(m,a,b,0.618,5);
	cout << tmp[0] << tmp[1] << endl;
	int finalPosition = HillClimbingSearch(m, tmp[0], tmp[1]);
	cout << finalPosition << endl;
	system("pause");
	return 0;
}

//黄金分割粗搜索阶段
/*
参数:已经确定的最大可调范围【a，b】；
一个脉冲移动1um；一个脉冲距离l；
黄金分割比例k=0.618；
已经确定的区间阈值T；
返回值：一个区间 vector（）
*/
vector<int> GoldenSectionSearch(double m[], int a, int b, double k, int T)
{
	vector<int> range;
	if (abs(b - a) < T)
	{
		range.push_back(a);
		range.push_back(b);
		return range;
	}

	int c = b - k*(b - a);
	int d = a + k*(b - a);

	if (m[c] > m[d])
	{
		b = d;
		return GoldenSectionSearch(m, a, b, k, T);
	}
	else
	{
		a = c;
		return GoldenSectionSearch(m, a, b, k, T);
	}

	return range;
}

//爬山精细化搜索阶段
/*
参数：区间【c，d】；
最小脉冲数s；
*/
int HillClimbingSearch(double m[], int c, int d)
{
	double max = m[c];
	for (int i = 0; i < abs(c - d); i++)
	{
		double tmp = m[++c];
		if (tmp < max)
		{
			return --c;
		}
		max = tmp;
	}
	return -1;
}

//图像评价函数--方差函数--Halocn转C++
int EvaluationImage(int k)
{
	// Local iconic variables
	HObject  ho_Image;
	HObject  ho_GrayImage, ho_Regions, ho_RegionTrans, ho_ROI;
	HObject  ho_LU, ho_LD, ho_RU;

	// Local control variables
	HTuple  hv_Width, hv_Height, hv_Row, hv_Column;
	HTuple  hv_Phi, hv_Length1, hv_Length2, hv_Mean, hv_Deviation;

	ReadImage(&ho_Image, "F:/7号屏/W/20200923（较虚焦，实验平台，VP101M）-7号屏-5-W255-曝光时间20ms-OPT平面光源（10）-国产镜头（光圈5.6）.bmp");
	Rgb1ToGray(ho_Image, &ho_GrayImage);
	GetImageSize(ho_GrayImage, &hv_Width, &hv_Height);
	Threshold(ho_GrayImage, &ho_Regions, 50, 80);
	ShapeTrans(ho_Regions, &ho_RegionTrans, "convex");
	SmallestRectangle2(ho_RegionTrans, &hv_Row, &hv_Column, &hv_Phi, &hv_Length1, &hv_Length2);
	GenRectangle2(&ho_ROI, hv_Row, hv_Column, hv_Phi, hv_Length1 / 5, hv_Length2 / 5);
	GenRectangle2(&ho_LU, hv_Row - (0.6*hv_Length2), hv_Column - (0.6*hv_Length1), hv_Phi,hv_Length1 / 5, hv_Length2 / 5);
	Union2(ho_ROI, ho_LU, &ho_ROI);
	GenRectangle2(&ho_LD, hv_Row + (0.6*hv_Length2), hv_Column - (0.6*hv_Length1), hv_Phi,hv_Length1 / 5, hv_Length2 / 5);
	Union2(ho_ROI, ho_LD, &ho_ROI);
	GenRectangle2(&ho_RU, hv_Row - (0.6*hv_Length2), hv_Column + (0.6*hv_Length1), hv_Phi,hv_Length1 / 5, hv_Length2 / 5);
	Union2(ho_ROI, ho_RU, &ho_ROI);
	GenRectangle2(&ho_LU, hv_Row + (0.6*hv_Length2), hv_Column + (0.6*hv_Length1), hv_Phi,hv_Length1 / 5, hv_Length2 / 5);
	Union2(ho_ROI, ho_LU, &ho_ROI);
	Intensity(ho_ROI, ho_GrayImage, &hv_Mean, &hv_Deviation);
	return hv_Deviation.I();
}