算法描述
SIFT特征不只具有尺度不变性，即使改变旋转角度，图像亮度或拍摄视角，仍然能够得到好的检测效果。整个算法分为以下几个部分：
（1）构造尺度空间：DoG尺度空间
（2）检测DoG尺度空间极值点
（3）去除不好的特征点
（4）给特征点赋值一个128维的方向参数。每个关键点都包含三个信息：位置、尺度和方向。
（5）关键点描述子的生成：
首先将坐标轴旋转为关键点的方向，以确保旋转不变性。以关键点为中心取8×8的窗口。
（6）最后进行特征匹配。
Algorithm description SIFT features not only have scale invariance, but also can obtain good detection results even if the rotation angle, image brightness or shooting angle are detected. The whole algorithm is divided into the following parts: (1) tectonic scale space: DoG scale space (2) detecting the spatial extreme points of DoG scale (3) remove bad feature points (4) assign a 128 dimensional direction parameter to the feature point. Each key point contains three information: location, scale and direction. (5) generation of key point descriptors: Firstly, the coordinate axis is rotated as the direction of the key point to ensure the rotation invariance. Take the key point as the center and take the window of 8 x 8. (6) feature matching is performed at last. (2017-10-22, C++, 1154KB, 下载2次)
用root程序读取ntuple文件的信息时(root -l root_script.cxx)，需要此类型的画图脚本。输入存储信息的.root文件，取出指定tree下的leaf信息，画在一维直方图上，调整cut，不同的输入间图形对比，看图形形状的变化是否合理，进而解决问题。注：本程序在相关领域较实用，基本的东西都有，易于理解，借此入门，举一反三。
When reading information ntuple program file with the root (root-l root_script.cxx), you require this type of drawing scripts. Enter to store information .root file, remove the leaf information specified under the tree, painted on a one-dimensional histogram, adjust the cut, the graphical comparison between different input to see whether reasonable changes in pattern shape, and then solve the problem. Note: This program more practical in related fields, the basic things are easy to understand, thereby getting started, learn by analogy. (2016-05-06, C++, 1KB, 下载1次)