【机器学习】改善LBP特征提高SVM的可分性的经验总结（一）

1、等价模式LBP的最初效果

正样本2343，负样本5883，直接计算等价模式LBP特征图，再使用SVM训练一个分类器。

CvTermCriteria criteria = cvTermCriteria(CV_TERMCRIT_ITER, 1000, FLT_EPSILON);

CvSVMParams param(CvSVM::C_SVC, CvSVM::LINEAR, 0, 0.5, 0, 0.01, 0, 0, 0, criteria);

 

2、添加高斯滤波处理

GaussianBlur(inputImg, inputImg, Size(3,3), 0);

结论：加入高斯滤波，由于可去除噪声干扰；车道线检出率提高，多检率降低。

 

3、再把LBP特征计算的像素差值增大至5 

// LBP特征的实现及LBP+SVM分类 https://blog.csdn.net/qianqing13579/article/details/49406563
// 计算等价模式LBP特征图，为了方便表示特征图，58种等价模式表示为1~58,第59种混合模式表示为0
static void ComputeLBPImage_Uniform(const Mat &srcImage, Mat &LBPImage)
{
	// 参数检查，内存分配
	CV_Assert(srcImage.depth() == CV_8U&&srcImage.channels() == 1);
	LBPImage.create(srcImage.size(), srcImage.type());

	// 计算LBP图
	// 扩充原图像边界，便于边界处理
	Mat extendedImage;
	copyMakeBorder(srcImage, extendedImage, 1, 1, 1, 1, BORDER_DEFAULT);

	// 构建LBP 等价模式查找表
	//int table[256];
	//BuildUniformPatternTable(table);

	// LUT(256种每一种模式对应的等价模式)
	static const int table[256] = { 1, 2, 3, 4, 5, 0, 6, 7, 8, 0, 0, 0, 9, 0, 10, 11, 12, 0, 0, 0, 0, 0, 0, 0, 13, 0, 0, 0, 14, 0, 15, 16, 17, 0, 0, 0,
		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 18, 0, 0, 0, 0, 0, 0, 0, 19, 0, 0, 0, 20, 0, 21, 22, 23, 0, 0, 0, 0, 0, 0, 0, 0,
		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 24, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 25,
		0, 0, 0, 0, 0, 0, 0, 26, 0, 0, 0, 27, 0, 28, 29, 30, 31, 0, 32, 0, 0, 0, 33, 0, 0, 0, 0, 0, 0, 0, 34, 0, 0, 0, 0
		, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 35, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
		0, 0, 0, 0, 36, 37, 38, 0, 39, 0, 0, 0, 40, 0, 0, 0, 0, 0, 0, 0, 41, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 42
		, 43, 44, 0, 45, 0, 0, 0, 46, 0, 0, 0, 0, 0, 0, 0, 47, 48, 49, 0, 50, 0, 0, 0, 51, 52, 53, 0, 54, 55, 56, 57, 58 };

	// 计算LBP
	int heightOfExtendedImage = extendedImage.rows;
	int widthOfExtendedImage = extendedImage.cols;
	int widthOfLBP=LBPImage.cols;
	uchar *rowOfExtendedImage = extendedImage.data+widthOfExtendedImage+1;
	uchar *rowOfLBPImage = LBPImage.data;

	int pixelDiff = 5;

	for (int y = 1; y <= heightOfExtendedImage - 2; ++y,rowOfExtendedImage += widthOfExtendedImage, rowOfLBPImage += widthOfLBP)
	{
		// 列
		uchar *colOfExtendedImage = rowOfExtendedImage;
		uchar *colOfLBPImage = rowOfLBPImage;
		for (int x = 1; x <= widthOfExtendedImage - 2; ++x, ++colOfExtendedImage, ++colOfLBPImage)
		{
			// 计算LBP值
			int LBPValue = 0;
			if (colOfExtendedImage[0 - widthOfExtendedImage - 1] >= colOfExtendedImage[0]+pixelDiff)
				LBPValue += 128;
			if (colOfExtendedImage[0 - widthOfExtendedImage] >= colOfExtendedImage[0]+pixelDiff)
				LBPValue += 64;
			if (colOfExtendedImage[0 - widthOfExtendedImage + 1] >= colOfExtendedImage[0]+pixelDiff)
				LBPValue += 32;
			if (colOfExtendedImage[0 + 1] >= colOfExtendedImage[0]+pixelDiff)
				LBPValue += 16;
			if (colOfExtendedImage[0 + widthOfExtendedImage + 1] >= colOfExtendedImage[0]+pixelDiff)
				LBPValue += 8;
			if (colOfExtendedImage[0 + widthOfExtendedImage] >= colOfExtendedImage[0]+pixelDiff)
				LBPValue += 4;
			if (colOfExtendedImage[0 + widthOfExtendedImage - 1] >= colOfExtendedImage[0]+pixelDiff)
				LBPValue += 2;
			if (colOfExtendedImage[0 - 1] >= colOfExtendedImage[0]+pixelDiff)
				LBPValue += 1;

			colOfLBPImage[0] = table[LBPValue];
		} // x

	}// y
}

结论：计算LBP编码值，像素差值取5，可以减少背景细微差异而又满足车道线特征分布带来的多检，整体而言检测结果更加准确。