简体繁体 English

识别多张面孔（eigenfaces opencv）

[英]Recognizing multiple faces (eigenfaces opencv)

原文 2014-01-13 15:00:34 4 1 c++/ opencv/ mfc

I'm creating facial recognition program that uses eigenfaces method implemented in OpenCV 2.4.8. 我正在创建使用OpenCV 2.4.8中实现的eigenfaces方法的面部识别程序。 I have training set of three faces (10 pictures each), also i have 3 pictures of people trained in training set. 我有三张脸的训练集（每张10张图片），也有3人训练集的图片。 So far program works very well, recognition is 100% correct. 到目前为止，程序运行良好，识别是100％正确的。 However, I have tried to solve different problem with my program.. I have edited 3 test photos so that i have put those 3 photos in one photo, and that photo is my test sample. 但是，我尝试用我的程序解决其他问题。.我编辑了3张测试照片，以便将这3张照片放在一张照片中，并且该照片是我的测试样本。 Program doesn't recognize anyone in that example. 在该示例中，程序无法识别任何人。

So my conclusion would be that eigenfaces recognizer can recognize only one face per photo. 所以我的结论是，特征脸识别器每张照片只能识别一张脸。 Is that correct? 那是对的吗？

Is there some kind of eignefaces support to recognize multiple faces in one photo? 是否存在某种eignefaces支持来识别一张照片中的多张面孔？

1 个解决方案

OpenCV face recognition can only work on a face at a time. OpenCV人脸识别一次只能在人脸上使用。 Your program cannot recognize anyone from the test image since the euclidean distance it computes from the test image and your training set is simply too high. 您的程序无法从测试图像中识别出任何人，因为它从测试图像计算出的欧式距离以及您的训练集太高了。 You would need to implement a face detection algorithm if you want to test an image with several faces. 如果要测试具有多个面部的图像，则需要实现面部检测算法。