首先看看get_features函数。

            首先判断是hog特征还是gray,分两种情况。

            如果是hog特征,调用fhog函数,返回x,并将矩阵x的第三维最后一个组数据删除(好奇fhog函数http://vision.ucsd.edu/~pdollar/toolbox/doc/index.html)。

            

  1. if features.hog,
  2. %HOG features, from Piotr's Toolbox
  3. x = double(fhog(single(im) / 255, cell_size, features.hog_orientations));
  4. x(:,:,end) = []; %remove all-zeros channel ("truncation feature")
  5. %将矩阵x的第三维最后一个组数据删除
  6. end

            如果是gray,则直接对其归一化处理。

             

  1. if features.gray,
  2. %gray-level (scalar feature)
  3. x = double(im) / 255;
  4.  
  5. x = x - mean(x(:));
  6. end

函数最后返回x

    下面分析下各个核的correlation。

        首先是gaussian核,计算跟CSK没什么两样:

        

  1. N = size(xf,1) * size(xf,2);
  2. xx = xf(:)' * xf(:) / N; %squared norm of x
  3. yy = yf(:)' * yf(:) / N; %squared norm of y
  4. %cross-correlation term in Fourier domain
  5. xyf = xf .* conj(yf);
  6. xy = sum(real(ifft2(xyf)), 3); %to spatial domain
  7. %calculate gaussian response for all positions, then go back to the
  8. %Fourier domain
  9. kf = fft2(exp(-1 / sigma^2 * max(0, (xx + yy - 2 * xy) / numel(xf))));

        

   对于linear kernel

  1. kf = sum(xf .* conj(yf), 3) / numel(xf);

    对于polynomial kernel

  1. 1 xyf = xf .* conj(yf);
  2. xy = sum(real(ifft2(xyf)), 3); %to spatial domain
  3. %calculate polynomial response for all positions, then go back to the
  4. %Fourier domain
  5. kf = fft2((xy / numel(xf) + a) .^ b);

High-Speed Tracking with Kernelized Correlation Filters(二)的更多相关文章

  1. KCF:High-Speed Tracking with Kernelized Correlation Filters 的翻译与分析(一)。分享与转发请注明出处-作者:行于此路

    High-Speed Tracking with Kernelized Correlation Filters 的翻译与分析 基于核相关滤波器的高速目标跟踪方法,简称KCF 写在前面,之所以对这篇文章 ...

  2. High-Speed Tracking with Kernelized Correlation Filters

          2015年的一篇论文,可参考:http://blog.csdn.net/carrierlxksuper/article/details/46461245.      另参考:http:// ...

  3. Correlation Filter in Visual Tracking系列一:Visual Object Tracking using Adaptive Correlation Filters 论文笔记

    Visual Object Tracking using Adaptive Correlation Filters 一文发表于2010的CVPR上,是笔者所知的第一篇将correlation filt ...

  4. Hign-Speed Tracking with Kernelzied Correlation Filters

    reference:Hign-Speed Tracking with Kernelzied Correlation Filters questions: The core componet of mo ...

  5. call Kernelized Correlation Filters Tracker(Matab) in Qt(c++)

    recently, i need call the KCF tracker in my graduation project. the KCF tracker is fast and best per ...

  6. correlation filters in object tracking

    http://www.cnblogs.com/hanhuili/p/4266990.html Correlation Filter in Visual Tracking系列一:Visual Objec ...

  7. Learning Spatial-Temporal Regularized Correlation Filters for Visual Tracking---随笔

    Learning Spatial-Temporal Regularized Correlation Filters for Visual Tracking DCF跟踪算法因边界效应,鲁棒性较差.SRD ...

  8. Multi-hierarchical Independent Correlation Filters for Visual Tracking(MFT)略读

    作者写道: 有幸在本届的VOT 2018 主赛中,我们的参赛方案Multi-solution Fusion for Visual Tracking(MFT)获得第一名的成绩,通过结果来看,MFT无论在 ...

  9. correlation filters in object tracking2

    http://www.cnblogs.com/hanhuili/p/4281077.html Correlation Filter in Visual Tracking系列二:Fast Visual ...

随机推荐

  1. Andorid APK反逆向

    Andorid APK反逆向解决方案---梆梆加固原理探寻http://blog.csdn.net/androidsecurity/article/details/8892635 Android AP ...

  2. GO_05_2:Golang 中 panic、recover、defer 的用法

     函数 defer 1. 它的执行方式类似其他语言中的折构函数,在函数体执行结束后按照调用顺序的 相反顺序 逐个执行 2. 即使函数发生 严重错误 也会被执行,类似于 java 中 try{...} ...

  3. ROI POOLING 介绍

    转自 https://blog.csdn.net/gbyy42299/article/details/80352418 Faster rcnn的整体构架: 训练的大致过程: 1.图片先缩放到MxN的尺 ...

  4. NAT—网络地址转换

    参考链接:http://www.qingsword.com/qing/745.html 视频链接: 一.什么是NAT? NAT --- Network Address Translation  也就是 ...

  5. Ubuntu16.04.2安装Tensorflow

    安装aptitude $ sudo apt-get install aptitude 安装python-pip python-dev $ sudo aptitude install python-pi ...

  6. bzoj千题计划146:bzoj3295: [Cqoi2011]动态逆序对

    http://www.lydsy.com/JudgeOnline/problem.php?id=3295 正着删除看做倒着添加 对答案有贡献的数对满足以下3个条件: 出现时间:i<=j 权值大小 ...

  7. 2017 清北济南考前刷题Day 7 morning

    期望得分:100+50+20=170 实际得分:10+50+20=80 1. 纸牌 题目描述 在桌面上放着n张纸牌,每张纸牌有两面,每面都写着一个非负整数.你的邪王真眼可以看到所有牌朝上的一面和朝下的 ...

  8. An Insight to References in C++

    [An Insight to References in C++] 引用的本质是常指针.占用的内存和指针一样. 参考:http://www.codeproject.com/Articles/13363 ...

  9. MySQL在net中Datatime转换

    <add name="adDb"         connectionString="Persist Security Info=False;database=ad ...

  10. [转载]Visual Studio支持Node.js

    http://news.cnblogs.com/n/193893/ https://nodejstools.codeplex.com/ 微软发布了一个官方插件“Node.js Tools for Vi ...