关于散点合拢的案例

需求：一推排列规律的散点图，将其尽可能的合拢


思路1.散点按照一定规律排列，故先矫正倾斜角度，在聚合点
思路2.直接对个点坐标进行统一缩小

执行过程发现思路1，在剔除空袭后再对放置各点，但是点于点之间的斜率关系被破坏，使得各点位置信息损失。

而思路2在这方面表现较好。

代码如下：

struct mc_and_pix
{
	vector<Point2f> mc;
	vector<vector<int>> pix_all;
};

//垂直方向投影
void VerticalProjection(Mat& srcImageBin, vector<int>& Vertical_singel)//垂直积分投影  
{
	int value;

	for (int i = 0; i < srcImageBin.cols; i++)
	{
		int icount = 0;
		for (int j = 0; j < srcImageBin.rows; j++)
		{
			value = srcImageBin.at<uchar>(j, i);
			if (value != 0)
			{
				icount++; //统计每列的非零像素点 
			}
		}
		if (0 == icount)
		{
			Vertical_singel[i] = 1;
		}
	}
}

void HorizontalProjection(Mat& srcImageBin, vector<int>& Horizontal_singel)
{
	int value;
	for (int i = 0; i < srcImageBin.rows; i++)
	{
		int icount = 0;
		for (int j = 0; j < srcImageBin.cols; j++)
		{
			value = srcImageBin.at<uchar>(i, j);
			if (value != 0)
			{
				icount++; //统计每行的非零像素点   	
			}

		}
		if (0 == icount)
		{
			Horizontal_singel[i] = 1;
		}
	}
}

//计算保留的行（列）数
int deal_R_C(vector<int>& P)
{
	int len = P.size();
	int k = 0;
	for (int i = 0; i < len; i++)
	{
		if (0 == P[i])
		{
			k++;
		}
	}
	return k;
}

//初步剔除原图中缝隙和背景
Mat first_deal(Mat& src1)
{
	vector<int> Vertical_s(src1.cols);
	vector<int> Horizontal_s(src1.rows);

	VerticalProjection(src1, Vertical_s);
	HorizontalProjection(src1, Horizontal_s);

	int d_col = deal_R_C(Vertical_s);
	int d_row = deal_R_C(Horizontal_s);

	Mat tmp_r = Mat::zeros(Size(src1.cols, d_row), CV_8UC1);
	int b = 0;
	for (int j = 0; j < src1.rows; j++)
	{
		if (0 == Horizontal_s[j])
		{
			src1.row(j).copyTo(tmp_r.row(b));
			b++;
		}
	}

	Mat tmp = Mat::zeros(Size(d_col, d_row), CV_8UC1);
	int p = 0;
	for (int i = 0; i < tmp_r.cols; i++)
	{
		if (0 == Vertical_s[i])
		{
			tmp_r.col(i).copyTo(tmp.col(p));
			p++;
		}
	}
	return tmp;
}

//获取中心 并将附近 5*5 区域储存起来
void computer_mu(Mat& src1, Mat& src2, mc_and_pix& MC)
{
	vector<vector<Point>> contours;
	findContours(src2, contours, RETR_EXTERNAL, CHAIN_APPROX_NONE);

	/*Mat dd = Mat::zeros(Size(src.cols, src.rows), CV_8UC1);
	drawContours(dd, contours, 0, Scalar(255), 5, 8);*/

	/// 计算矩
	vector<Moments> mu(contours.size());
	for (int i = 0; i < contours.size(); i++)
	{
		mu[i] = moments(contours[i], false);
	}
	///  计算中心矩:
	vector<Point2f> mc(contours.size());
	for (int i = 0; i < contours.size(); i++)
	{
		mc[i] = Point2f(mu[i].m10 / mu[i].m00, mu[i].m01 / mu[i].m00);
	}

	uchar* s_data = src1.data;
	int num = mc.size();
	vector<vector<int>> pix_all(num, vector<int>(25));

	for (int m = 0; m < num; m++)
	{
		for (int j = -2; j < 3; j++)
		{
			for (int i = -2; i < 3; i++)
			{
				int x = int(mc[m].x) + i;
				int y = int(mc[m].y) + j;
				int pix = s_data[y * src1.step + x];
				int k = (j + 2) * 5 + (i + 2);
				pix_all[m][k] = pix;
			}
		}
	}

	MC.mc = mc;
	MC.pix_all = pix_all;
}

int main()
{
	Mat src = imread("img1.bmp", 0);

	Mat tmp;
	tmp = first_deal(src);

	copyMakeBorder(tmp, tmp, 3, 3, 3, 3, BORDER_CONSTANT, Scalar(0));

	Mat src1;
	src1 = tmp > 3;

	Mat src2;
	Mat kernel1 = getStructuringElement(MORPH_RECT, Size(5, 5));
	dilate(src1, src2, kernel1);

	mc_and_pix MC;
	computer_mu(tmp, src2,MC);	 

	Mat dst_roi_1 = Mat::zeros(Size(src2.cols, src2.rows), CV_8UC1);
	uchar* d_r_1 = dst_roi_1.data;

	for (int mm = 0; mm < MC.mc.size(); mm++)
	{
		int gx = MC.mc[mm].x * 0.2;
		int gy = MC.mc[mm].y * 0.2;

		for (int m = -2; m < 3; m++)
		{
			for (int n = -2; n < 3; n++)
			{
				d_r_1[(gy + m + 100) * src2.step + (gx + n + 100)] = MC.pix_all[mm][5 * (m + 2) + (n + 2)];
			}
		}
	}
	waitKey(0);
	return 0;
}