最近在看视觉跟踪方面的论文,ZK博士的TLD算法作为跟踪算法的state-of-the-art,当然不得不去拜读下了,看完论文后虽然对作者整体的思想有了一个大致了解,但是对于很多细节却也还是无从得知,好在作者将自己的算法源代码全部开源,这也造福了我们这些无知者的胃口,虽然网上有几个c++版本的源码,但是matlab版本作为作者的原始版本,拿来作研究也是极好的。
通览源代码,个人感觉精髓之处无非两个函数,即:
tldInit();
tldProcessFrame();
这二者一个是用来跟踪前的初始化工作,后者当然是run-time的跟踪工作。我个人的阅读注释也将从这两个函数展开。好了,闲话少叙,下面来看看tldInit()这个函数:
进去第一个函数是
- %输入:
- %tld.source.bb 用户输入的目标标定框
- %size(tld.source.im0.input) 输入图像的尺寸
- %tld.model.min_win 目标标定框的长或宽的最小尺寸
- %输出:
- %tld.grid 是一个6Xn(n表示不同有效尺度下共有多少个的网格)的矩阵,前四行组成列向量表示gridbox的4个顶点,5表示索引,6表示横向的分布点数
- % tld.scales 有效尺度下gridbox的高和宽,为2Xm矩阵(m表示有效的尺度的个数)
- [tld.grid tld.scales] = bb_scan(tld.source.bb,size(tld.source.im0.input),tld.model.min_win);
再让我们继续进入bb_scan()函数:
- bbW = round(bb_width(bb) .* SCALE);%bb_width(bb)*(0.1615~6.1917)矩形框变尺度
- bbH = round(bb_height(bb) .* SCALE);
- bbSHH = SHIFT * min(bbH,bbH);%取bbH和bbH中最小的元素组成新的矩阵,维数和原来相同 <span style="font-family: Arial, Helvetica, sans-serif;">SHIFT =</span>0.1表示变形后的矩形框的移动步长
- bbSHW = SHIFT * min(bbH,bbW);
- bbF = [2 2 imsize(2) imsize(1)]';%[2 2 320 240]
- bbs = {};
- sca = [];
- idx = 1;
- for i = 1:length(SCALE)
- if bbW(i) < MINBB || bbH(i) < MINBB, continue; end
- left = round(bbF(1):bbSHW(i):bbF(3)-bbW(i)-1);%2开头,0.1* min(bbH(i),bbW(i))为步长,320- bbW(i)-1为结尾
- top = round(bbF(2):bbSHH(i):bbF(4)-bbH(i)-1);%2开头,0.1* min(bbH(i),bbH(i))为步长,240- bbH(i)-1为结尾
- grid = ntuples(top,left);%重复功能,维度为2 X length(top)*length(left);grid的每个列向量成为网格顶点的坐标点
- if isempty(grid), continue; end
- bbs{end+1} = [grid(2,:); ...%省略号为续行,bbs前四行的列向量构成一个矩形网格的四个顶点
- grid(1,:); ...
- grid(2,:)+bbW(i)-1; ...
- grid(1,:)+bbH(i)-1; ...
- idx*ones(1,size(grid,2));%坐标点对应的编号,从1开始
- length(left)*ones(1,size(grid,2));];%记录横向分布多少个点?
- sca = [sca [bbH(i); bbW(i)]];%在原来SCA的基础上增加[bbH(i); bbW(i)]列向量,该列向量表示某个尺度因子下的gridbox的高和宽
- idx = idx + 1;
- end
- bb_out = [];
- for i = 1:length(bbs)
- bb_out = [bb_out bbs{i}];
- end
下面来到了产生特征的函数
- %输入:
- %tld.model.num_trees:有10棵分类树
- %tld.model.num_features:每棵树里有13个特性,这里是会有13个点对的比较
- %输出:
- %tld.features: 4*13X10的矩阵,为fern分类器选取的随机点对,4表示一个点对的四个坐标值
- tld.features = tldGenerateFeatures(tld.model.num_trees,tld.model.num_features,1);
- SHI = 1/5;
- SCA = 1;
- OFF = SHI;
- x = repmat(ntuples(0:SHI:1,0:SHI:1),2,1);%ntuples生成网格点坐标2X36矩阵,repmat表示将生成的矩阵作为初始化元素(还是矩阵)
- %生成一个2X1的大矩阵,这里是4X36(6*6)矩阵
- x = [x x + SHI/2];%前者4X36矩阵,后者矩阵偏移0.1
- k = size(x,2);%k = 36*2
- %rand(1,k)随机生成都1Xk维随机分布矩阵,范围0~1
- r = x; r(3,:) = r(3,:) + (SCA*rand(1,k)+OFF);%x的第3行加上随机的1X72矩阵(0~1)再加上0.2
- l = x; l(3,:) = l(3,:) - (SCA*rand(1,k)+OFF);%x的第3行减去随机的1X72矩阵(0~1)再减去0.2
- t = x; t(4,:) = t(4,:) - (SCA*rand(1,k)+OFF);%x的第4行减去随机的1X72矩阵(0~1)再减去0.2
- b = x; b(4,:) = b(4,:) + (SCA*rand(1,k)+OFF);%x的第4行加上随机的1X72矩阵(0~1)再加上0.2
- x = [r l t b];
- idx = all(x([1 2],:) < 1 & x([1 2],:) > 0,1);%idx和X维数相同,如果满足条件相应元素为1,否则为0
- x = x(:,idx);%挑选满足条件的所有X的列
- x(x > 1) = 1;%x元素中大于1的则用1代替此元素
- x(x < 0) = 0;%x元素中小于0的则用0代替此元素
- numF = size(x,2);%看看现在X还有多少列
- x = x(:,randperm(numF));%randperm(numF)从1~numF的数字序列随机打乱,整个表达式是指把X的列打乱
- x = x(:,1:nFEAT*nTREES);%取x的前130列,x目前为4x130矩阵
- x = reshape(x,4*nFEAT,nTREES);%把X塑造成4*13 X 10 的矩阵
下面再来到fern(0);没什么好说的,来看下面的吧:
- %输入:
- %tld.source.im0.input:输入的图像
- %tld.grid:6Xn(n表示不同有效尺度下共有多少个的网格)的矩阵,gridbox信息
- %tld.features:4*13X10的矩阵,为fern分类器选取的随机点对,4表示一个点对的四个坐标值
- %tld.scales: 有效尺度下gridbox的高和宽,为2Xm矩阵(m表示有效的尺度的个数)
- %输出:
- fern(1,tld.source.im0.input,tld.grid,tld.features,tld.scales); % allocate structures
- iHEIGHT = mxGetM(prhs[1]);//240
- iWIDTH = mxGetN(prhs[1]);//320
- nTREES = mxGetN(mxGetField(prhs[3],0,"x"));//<span style="font-family: Arial, Helvetica, sans-serif;">10</span>
- nFEAT = mxGetM(mxGetField(prhs[3],0,"x")) / 4; // feature has 2 points: x1,y1,x2,y2 13
- thrN = 0.5 * nTREES;//等于5
- nSCALE = mxGetN(prhs[4]);//获得共有多少种尺度的BOX,即有效尺度的个数
- IIMG = (double*) malloc(iHEIGHT*iWIDTH*sizeof(double));//积分图像变量准备
- IIMG2 = (double*) malloc(iHEIGHT*iWIDTH*sizeof(double));//平方积分图像变量准备
- // BBOX
- mBBOX = mxGetM(prhs[2]);//等于6
- nBBOX = mxGetN(prhs[2]);//等于在各个有效尺度下有多少个网格
- BBOX = create_offsets_bbox(mxGetPr(prhs[2]));//创建保存网格数据索引等数据
- //prhs[3]是4*13X10的矩阵
- //matlab 代码中有f.x = x;f.type = 'forest';见到下面就不怪了
- double *x = mxGetPr(mxGetField(prhs[3],0,"x"));//获得特征点的指针
- double *s = mxGetPr(prhs[4]);//各种尺度的BOX的尺寸
- OFF = create_offsets(s,x);//记录各个特征点对在各种尺度下box中的具体位置
- for (int i = 0; i<nTREES; i++) {
- WEIGHT.push_back(vector<double>(pow(2.0,nBIT*nFEAT), 0));//nBIT=1,权重分配权2^13
- nP.push_back(vector<int>(pow(2.0,nBIT*nFEAT), 0));//nBIT=1,nP分配2^13
- nN.push_back(vector<int>(pow(2.0,nBIT*nFEAT), 0));//nBIT=1,nN分配2^13
- }
- //static vector<vector <double> > WEIGHT;
- //static vector<vector <int> > nP;
- //static vector<vector <int> > nN;下面见怪不怪,10X 2^13的容器
- for (int i = 0; i<nTREES; i++) {
- for (int j = 0; j < WEIGHT[i].size(); j++) {
- WEIGHT[i].at(j) = 0;
- nP[i].at(j) = 0;
- nN[i].at(j) = 0;
- }
- }
来具体看看create_offset_bbox和creater_offset两个函数
- int *offsets = (int*) malloc(BBOX_STEP*nBBOX*sizeof(int));//7*nBBOX*sizeof(int)
- int *off = offsets;
- for (int i = 0; i < nBBOX; i++) {//nBBOX表示所有的网格数
- double *bb = bb0+mBBOX*i;//偏移到下一个网格的属性的向量
- //sub2idx(row,col,height) ((int) (floor((row)+0.5) + floor((col)+0.5)*(height)))
- //floor不大于
- //bb[0]:left bb[1]top bb[2]right bb[3]bottom iHEIGHT:240 bb[4]索引从1开始 bb[5]表示number of left-right bboxes
- //以下记录索引是从左到右,从上到下方式
- *off++ = sub2idx(bb[1]-1,bb[0]-1,iHEIGHT);//左上顶点索引
- *off++ = sub2idx(bb[3]-1,bb[0]-1,iHEIGHT);//左下顶点索引
- *off++ = sub2idx(bb[1]-1,bb[2]-1,iHEIGHT);//右上顶点索引
- *off++ = sub2idx(bb[3]-1,bb[2]-1,iHEIGHT);//右下顶点索引
- *off++ = (int) ((bb[2]-bb[0])*(bb[3]-bb[1]));//记录当前网格的大小
- *off++ = (int) (bb[4]-1)*2*nFEAT*nTREES; // pointer to features for this scale
- *off++ = bb[5]; // number of left-right bboxes, will be used for searching neighbours
- }
- return offsets;
- int *offsets = (int*) malloc(nSCALE*nTREES*nFEAT*2*sizeof(int));//
- int *off = offsets;
- for (int k = 0; k < nSCALE; k++){//共有多少种尺度的BOX
- double *scale = scale0+2*k;//scale0第一种尺度的尺寸信息,*2表示列向量有2维,即高和宽,此表达式表示偏移到下一个box的尺寸向量
- for (int i = 0; i < nTREES; i++) {//10
- for (int j = 0; j < nFEAT; j++) {//13
- //x0 4*13 X 10 的矩阵
- double *x = x0 +4*j + (4*nFEAT)*i;//4*j因为每个feature是一个4维列向量,(4*13)*i即下一棵树
- //sub2idx(row,col,height) ((int) (floor((row)+0.5) + floor((col)+0.5)*(height)))
- //scale[1]宽,scale[0]高,x[0]x坐标,x[1]y坐标
- *off++ = sub2idx((scale[0]-1)*x[1],(scale[1]-1)*x[0],iHEIGHT);//记录第一个点在该尺度BOX的具体位置,并转化为索引
- *off++ = sub2idx((scale[0]-1)*x[3],(scale[1]-1)*x[2],iHEIGHT);//记录第二个点在该尺度BOX的具体位置,并转化为索引
- }
- }
- }
今天继续,下面是开始要生成正负例来训练分类器了,首先:
- // TRAIN DETECTOR ==========================================================
- // Initialize structures
- tld.imgsize = size(tld.source.im0.input);
- //为fern准备的训练集
- tld.X = cell(1,length(tld.source.idx)); //training data for fern
- tld.Y = cell(1,length(tld.source.idx));
- %为nearest neighbor准备的训练集
- tld.pEx = cell(1,length(tld.source.idx)); // training data for NN
- tld.nEx = cell(1,length(tld.source.idx));
- //输入:
- //tld.source.bb:用户目标标定框
- //tld.grid: 生成的gridbox信息矩阵
- //输出:
- // overlap一维行向量,记录GRID中的各个gridbox与用户目标标定框的重叠率
- overlap = bb_overlap(tld.source.bb,tld.grid);
- // Input
- double *bb1 = mxGetPr(prhs[0]); int M1 = mxGetM(prhs[0]); int N1 = mxGetN(prhs[0]);//4X1
- double *bb2 = mxGetPr(prhs[1]); int M2 = mxGetM(prhs[1]); int N2 = mxGetN(prhs[1]);//6Xn(n表示gridbox总数)
- // Output
- if (N1 == 0 || N2 == 0) {
- N1 = 0; N2 = 0;
- }
- plhs[0] = mxCreateDoubleMatrix(N1, N2, mxREAL);//创建输出矩阵,1Xgridbox的数量
- double *out = mxGetPr(plhs[0]);
- for (int j = 0; j < N2; j++) {//gridbox的数量
- for (int i = 0; i < N1; i++) {//1
- *out++ = bb_overlap(bb1 + M1*i, bb2 + M2*j);//计算重叠度
- }
- }
- double bb_overlap(double *bb1, double *bb2) {
- if (bb1[0] > bb2[2]) { return 0.0; }//判断如果两个矩形没有相交部分,重叠度就为0;
- if (bb1[1] > bb2[3]) { return 0.0; }
- if (bb1[2] < bb2[0]) { return 0.0; }
- if (bb1[3] < bb2[1]) { return 0.0; }
- double colInt = min(bb1[2], bb2[2]) - max(bb1[0], bb2[0]) + 1;//求相交矩形的宽和高
- double rowInt = min(bb1[3], bb2[3]) - max(bb1[1], bb2[1]) + 1;
- double intersection = colInt * rowInt;//相交面积
- double area1 = (bb1[2]-bb1[0]+1)*(bb1[3]-bb1[1]+1);//分别求两个输入矩形的面积
- double area2 = (bb2[2]-bb2[0]+1)*(bb2[3]-bb2[1]+1);
- return intersection / (area1 + area2 - intersection);//求重叠率
- }
- //输入:
- //tld.img{1}.input:输入图像,第一帧
- //tld.bb(:,1):用户目标标定框
- //输出:
- //tld.target:目标标定框中特定的图像
- tld.target = img_patch(tld.img{1}.input,tld.bb(:,1));
- //如果4个坐标值都是整数
- if sum(abs(round(bb)-bb))==0
- L = max([1 bb(1)]);
- T = max([1 bb(2)]);
- R = min([size(img,2) bb(3)]);
- B = min([size(img,1) bb(4)]);
- patch = img(T:B,L:R);//在不超过画面尺寸和小于1x1的情况下,取出BB框出的画面
- % Sub-pixel accuracy
- else
- cp = 0.5 * [bb(1)+bb(3); bb(2)+bb(4)]-1;//bbox的中心坐标 center point
- %[1 0 -cp(1)]
- %[0 1 -cp(2)]
- %[0 0 1 ]
- H = [1 0 -cp(1); 0 1 -cp(2); 0 0 1];
- bbW = bb(3,:)-bb(1,:);//宽
- bbH = bb(4,:)-bb(2,:);//高
- if bbW <= 0 || bbH <= 0
- patch = [];
- return;
- end
- box = [-bbW/2 bbW/2 -bbH/2 bbH/2];
- if size(img,3) == 3//如果图像有三个通道,即判断图片是否为真彩色
- for i = 1:3
- P = warp(img(:,:,i),inv(H),box);
- patch(:,:,i) = uint8(P);
- end
- else
- patch = warp(img,inv(H),box);//inv(H)=[1 0 cp(1); 0 1 cp(2); 0 0 1];平移变换
- patch = uint8(patch);
- end
- end
好了下面开始产生正训练样本了:
- //输入:
- //overlap:一维行向量,记录GRID中的各个gridbox与用户目标标定框的重叠率
- //tld.p_par_init:opt.p_par_init= struct('num_closest',10,'num_warps',20,'noise',5,'angle',20,'shift',0.02,'scale',0.02);
- //输出:
- //pX:10 X length(idxP)*20 (length(idxP)<=10,20为'num_warps',20)的矩阵列向量表示一个gridbox的10棵树上的13位有效的code
- //pEx:225X1的列向量,各元素值为原像素值减去像素均值
- //bbP:最靠近BBOX的的gridbox,列向量表示该box的4个顶点
- [pX,pEx,bbP] = tldGeneratePositiveData(tld,overlap,tld.img{1},tld.p_par_init);
- pY = ones(1,size(pX,2));%1 X length(idxP)*20
- pX = [];
- pEx = [];
- // Get closest bbox
- [~,idxP] = max(overlap);//表示行不管,只取列,整个表达式表示最大overlap<span style="font-family: Arial, Helvetica, sans-serif;">所对应的列,一维</span>
- bbP0 = tld.grid(1:4,idxP);//1~4表示矩阵的4个顶点分布在四行,此取最靠近BBOX的的gridbox
- // Get overlapping bboxes
- idxP = find(overlap > 0.6);//返回overlap > 0.6所对应的列索引
- if length(idxP) > p_par.num_closest//如果overlap > 0.6的gridbox数大于10
- [~,sIdx] = sort(overlap(idxP),'descend'); //降序排序
- idxP = idxP(sIdx(1:p_par.num_closest));//取前p_par.num_closest个最大重叠度的bboxes所在的列
- end
- bbP = tld.grid(:,idxP);//取出10个最大重叠度的gridboxes
- if isempty(bbP), return; end
- % Get hull
- bbH = bb_hull(bbP);%得到能包围所有bbp中boxes<span style="font-family: Arial, Helvetica, sans-serif;">的最小矩形</span>
- cols = bbH(1):bbH(3);
- rows = bbH(2):bbH(4);
- im1 = im0;
- //返回一个225x1(pEx)的列向量,各元素值为原像素值减去像素均值
- pEx = tldGetPattern(im1,bbP0,tld.model.patchsize);//
- if tld.model.fliplr
- pEx = [pEx tldGetPattern(im1,bbP0,tld.model.patchsize,1)];
- end
- //返回20个正例
- for i = 1:p_par.num_warps//p_par.num_warps=20
- if i > 1
- randomize = rand; // Sets the internal randomizer to the same state
- //patch_input = img_patch(im0.input,bbH,randomize,p_par);
- //返回将画面进行仿射变换后的patch
- patch_blur = img_patch(im0.blur,bbH,randomize,p_par);//bbH包围所有bbp中bboxes的最小矩形
- //这个很重要,保证在C调用里的偏移的起始地址可以是一样的
- im1.blur(rows,cols) = patch_blur;//把仿射变换后的图像放到原图像对应的位置(<span style="font-family: Arial, Helvetica, sans-serif;">能包围所有bbp中boxes</span><span style="font-family: Arial, Helvetica, sans-serif;">的最小矩形</span>)
- //im1.input(rows,cols) = patch_input;
- end
- // Measures on blured image
- //单次返回10Xlength(idxP)的矩阵,列向量表示一个gridbox的10棵树上的13位code,
- //最后返回10Xlength(idxP)*20的矩阵
- pX = [pX fern(5,im1,idxP,0)];//idxP :overlap > 0.6所对应的列索引
- // Measures on input image
- //pEx(:,i) = tldGetPattern(im1,bbP0,tld.model.patchsize);
- //pEx = [pEx tldGetPattern(im1,tld.grid(1:4,idxP),tld.model.patchsize)];
- end
1.tldGetPattern()
- nBB = size(bb,2);//得到bbp0(最靠近BBOX的gridbox)的列,值为1
- pattern = zeros(prod(patchsize),nBB);//15*15 X 1 矩阵,返回矩阵
- if ~exist('flip','var')
- flip= 0;
- end
- // for every bounding box
- for i = 1:nBB//1
- // sample patch
- patch = img_patch(img.input,bb(:,i));//取出对应框中的图像
- // flip if needed
- if flip
- patch = fliplr(patch);
- end
- // normalize size to 'patchsize' and nomalize intensities to ZMUV
- //返回一个225x1的列向量,各元素值为原像素值减去像素均值
- pattern(:,i) = tldPatch2Pattern(patch,patchsize);//patch压缩变换到patchsize大小,然后将各个元素减去元素均值
- end
- patch = imresize(patch,patchsize); // 'bilinear' is faster
- pattern = double(patch(:));//此时变成225X1的矩阵
- pattern = pattern - mean(pattern);//mean(pattern)求各列向量的均值
- rand('state',randomize);
- randn('state',randomize);
- //'noise',5,'angle',20,'shift',0.02,'scale',0.02;
- NOISE = p_par.noise;
- ANGLE = p_par.angle;
- SCALE = p_par.scale;
- SHIFT = p_par.shift;
- cp = bb_center(bb)-1;//HULL矩形的中心
- Sh1 = [1 0 -cp(1); 0 1 -cp(2); 0 0 1];
- sca = 1-SCALE*(rand-0.5);%0.99~1.01
- //[0.99~1.01 ]
- //[ 0.99~1.01 ]
- //[ 1 ]
- Sca = diag([sca sca 1]);
- ang = 2*pi/360*ANGLE*(rand-0.5);//-10 ~ 10度 实际为弧度
- ca = cos(ang);
- sa = sin(ang);
- Ang = [ca, -sa; sa, ca];
- Ang(end+1,end+1) = 1;
- shR = SHIFT*bb_height(bb)*(rand-0.5);//-0.01~1.01*bb_height(bb)
- shC = SHIFT*bb_width(bb)*(rand-0.5);//-0.01~1.01*bb_width(bb)
- Sh2 = [1 0 shC; 0 1 shR; 0 0 1];
- bbW = bb_width(bb)-1;
- bbH = bb_height(bb)-1;
- box = [-bbW/2 bbW/2 -bbH/2 bbH/2];
- H = Sh2*Ang*Sca*Sh1;
- bbsize = bb_size(bb);
- patch = uint8(warp(img,inv(H),box) + NOISE*randn(bbsize(1),bbsize(2)));//给图像造成5的高斯噪声
3.fern()(第一个传参为5,获得模式)
- unsigned char *input = (unsigned char*) mxGetPr(mxGetField(prhs[1],0,"input"));
- unsigned char *blur = (unsigned char*) mxGetPr(mxGetField(prhs[1],0,"blur"));//获得仿射变换后的patch
- //if (mxGetM(prhs[1])!=iHEIGHT) { mexPrintf("fern: wrong input image.\n"); return; }
- // bbox indexes
- double *idx = mxGetPr(prhs[2]);//bbp所对应的列索引
- int numIdx = mxGetM(prhs[2]) * mxGetN(prhs[2]);//1 X (<=10)
- // minimal variance
- double minVar = *mxGetPr(prhs[3]);//minVar=0
- if (minVar > 0) {
- iimg(input,IIMG,iHEIGHT,iWIDTH);//返回IIMG,是图像进行矩形积分后的结果(运行不到这)
- iimg2(input,IIMG2,iHEIGHT,iWIDTH);//返回IIMG,是图像进行矩形平方积分后的结果(运行不到这)
- }
- // output patterns
- //创建输出矩阵:10X(<=10)
- plhs[0] = mxCreateDoubleMatrix(nTREES,numIdx,mxREAL);
- double *patt = mxGetPr(plhs[0]);
- //创建输出矩阵:1 X(<=10)
- plhs[1] = mxCreateDoubleMatrix(1,numIdx,mxREAL);
- double *status = mxGetPr(plhs[1]);
- for (int j = 0; j < numIdx; j++) {//(<=10)
- if (minVar > 0) {
- double bboxvar = bbox_var_offset(IIMG,IIMG2,BBOX+j*BBOX_STEP);//BBOX保存网格数据索引等数据(运行不到这)
- //E(p^2)-E^2(p)
- if (bboxvar < minVar) { continue; }(运行不到这)
- }
- status[j] = 1;
- double *tPatt = patt + j*nTREES;
- for (int i = 0; i < nTREES; i++) {//10
- //返回对应gridbox及对应树的13位有效的像素比较码
- tPatt[i] = (double) measure_tree_offset(blur, idx[j]-1, i);//idx:bbp
- }
- }
- return;
- int index = 0;
- int *bbox = BBOX + idx_bbox*BBOX_STEP;//BBOX存储gridbox的索引等信息BBOX_STEP=7(因为grid的行为6)
- //OFF + bbox[5],该表达式表示该gridbox的特征点信息在OFF的偏移,bbox[5]表示图像横向上多少个网格点
- //OFF = create_offsets(s,x);//记录各个特征点在各种尺度下box中的具体位置
- int *off = OFF + bbox[5] + idx_tree*2*nFEAT;//OFF存储特征点在各个尺度框下的分布位置等
- for (int i=0; i<nFEAT; i++) {//13
- index<<=1;
- //off[0]为特征点的x坐标,off[1]为特征点的y坐标,bbox[0]为该gridbox在图画中的位置
- int fp0 = img[off[0]+bbox[0]];
- int fp1 = img[off[1]+bbox[0]];
- if (fp0>fp1) { index |= 1;}//两个像素点比较并置位相应CODE
- off += 2;//移到下一个点对
- }
- return index;
- // Correct initial bbox
- tld.bb(:,1) = bbP(1:4,:);//最靠近BBOX的的gridbox
- // Variance threshold
- tld.var = var(pEx(:,1)) / 2;//var计算方差,这里即求各个数平方和的平均数
- // disp(['Variance : ' num2str(tld.var)]);
- // Generate Negative Examples
- //nx:patch variance 挑出合适的patches,并提取fern特征赋给nx,
- //nEx返回一个225x100(nEx)的矩阵,列向量各元素值为原像素值减去像素均值,100为num_patches
- //输入:
- //overlap:一维行向量,记录GRID中的各个gridbox与用户目标标定框的重叠率
- //输出:
- //nx:patch variance 挑出合适的patches,并提取fern特征赋给nx
- //nEx:一个225x100(nEx)的矩阵,列向量各元素值为原像素值减去像素均值,100为num_patches
- [nX,nEx] = tldGenerateNegativeData(tld,overlap,tld.img{1});
- // Measure patterns on all bboxes that are far from initial bbox
- //opt.n_par = struct('overlap',0.2,'num_patches',100);
- idxN = find(overlap<tld.n_par.overlap);//overlap < 0.2
- [nX,status] = fern(5,img,idxN,tld.var/2);//此函数通过patch variance剔除一批,剩下的进入fern特征码提取
- idxN = idxN(status==1); // bboxes far and with big variance,注意C++代码中的status[j] = 1;一句
- nX = nX(:,status==1);//选出进入第二级分类器的负样本
- // Randomly select 'num_patches' bboxes and measure patches
- idx = randvalues(1:length(idxN),tld.n_par.num_patches);//'num_patches',100应该是随机取出100个gridbox
- bb = tld.grid(:,idxN(idx));
- nEx = tldGetPattern(img,bb,tld.model.patchsize);//不复注解
好了,至此已经为分类器的训练产生了可用的正例和负例了。