基于深度学习的重叠人脸检测

人脸检测在人机界面、安全系统、人脸识别、基于内容的图像检索等不同应用中起着重要作用。随着计算机图像技术的发展,人脸检测的方法也越来越多。但是利用现有的人脸检测方法检测重叠人脸时,虽然能够检测出部分人脸,但是相比于单人脸的检测,算法的效率和准确性都有所欠缺。针对这个问题,提出了一种基于深度学习的重叠人脸检测方法。首先基于机器学习方法,构建出多个人脸特征分类器,然后再利用肤色检测的方法对分类器得到的候选人脸进行二次检测,最后利用提出的一种NMS算法对候选人脸进行进一步的处理,从而检测出精确的人脸。为了验证算法的高效性和准确性,进行了多个人脸检测算法的对比实验,结果表明,该算法在效率和准确性方面都有较大提高。

Overlapped Face Detection Based on Deep Learning

Face detection plays an important role in different applications such as human-machine interface,security system,face recognition and content-based image retrieval.With the development of computer image technology,there are more and more face detection methods.When detecting overlapping faces by the existing face detection methods,although partial faces can be detected,the efficiency and accuracy are insufficient compared to the detection of single faces.Aiming at this problem,we propose an overlapping face detection method based on deep learning.Firstly,based on the machine learning method,a number of face feature classifiers are constructed,and then the candidate face obtained by the classifier is secondarily detected by the skin color detection method.Finally,the candidate face is further processed by using an NMS algorithm proposed in this paper,so as to detect accurate faces.In order to verify the efficiency and accuracy of the algorithm,a comparative experiment of multiple face detection algorithms shows that the proposed algorithm has a great improvement in efficiency and accuracy.