出于安全目的的人脸识别

算法思想来自于网上资源，先使用图像边缘和车牌颜色定位车牌，再识别字符。车牌定位在predict方法中，为说明清楚，完成代码和测试后，加了很多注释，请参看源码。车牌字符识别也在predict方法中，请参看源码中的注释，需要说明的是，车牌字符识别使用的算法是opencv的SVM， opencv的SVM使用代码来自于opencv附带的sample，StatModel类和SVM类都是sample中的代码。SVM训练使用的训练样本来自于github上的EasyPR的c++版本。由于训练样本有限，你测试时会发现，车牌字符识别，可能存在误差，尤其是第一个中文字符出现的误差概率较大。(Bits in the predict method, for clarity, add a lot of comments after completing the code and testing, see the source code. License plate character recognition is also included in the predict method. Please refer to the comments in the source code. It needs to be noted that the algorithm used for license plate character recognition is SVM of opencv. The training sample used for SVM training comes from the c++ version of EasyPR on github. Due to limited training samples, you may find errors in license plate character recognition when testing, especially the error probability of the first Chinese character is large.)

说明：  SIFT特征点检测 DoG金字塔图像生成后，通过全搜索中间层DoG图像(去除第一个和最后一个DoG图像)的像素点，判断当前像素点的像素值是否大于(或者小于)周围8邻域的像素值(sddsdadadsda da da d ad a da adad ad a d ad a d ad a d ad a d)

说明：  MSER进行特征点提取，用的编译环境是vs2015(Feature point extraction by mser)

这是我做毕设时，用到的机器视觉关于OPENCV方面的文献，主要是尺寸测量方面的(This is what I do when the complete set, used in machine vision OPENCV the literature on the main aspects is the size measurement)

鱼眼校正代码，自动定标，可完美校正鱼眼的图片，已经跑完。

this function to load image from any folder using opencv

说明：  java 调用人脸识别，进行了封装，请解压内部文件，然后进行反编译jar(java call opencv Call face recognition, encapsulate, decompress internal files, and decompile jar)

包含了OpenCV2.4.3的使用手册和函数参考手册，非常适合编程时参考。(Contains a user manual and OpenCV2.4.3 function reference manual, very suitable for programming reference.)

对视频进行火焰检测，并实现对火焰的跟踪。(The video is detected by flame, and the flame is tracked)

opencv+MFC手势识别.使用OpenNI让Kinect识别出手势，并显示输出。目前版本的OpenNI支持四种手势：RaiseHand, Wave, Click和MovingHand，分别代表手的“举起”，“挥动”，“前推”和“移动”四种动作。值得一提的是，当前微软官方的Kinect SDK还不支持手势识别，也可以说是使用OpenNI的好处之一吧。