基于改进加速鲁棒特征的平面旋转人脸检测算法
首发时间:2017-02-28
摘要:针对人脸检测中存在的旋转人脸的问题,本文提出将改进的SURF(Speeded Up Robust Features)与Adaboost算法相结合并融合人眼定位来实现平面旋转人脸检测的方法。首先,对SURF算法进行改进,利用FAST算法代替Hessian矩阵来进行特征点的检测。将改进后的SURF算法与Adaboost算法相结合训练出一套SURF人脸公共特征集,然后提取图像的SURF特征并结合人眼定位来过滤无效特征点,最后将所提取的特征与人脸公共特征集进行匹配,从而实现平面旋转人脸的检测。实验结果表明:本文算法在检测率和误检率方面具有了较大的优势,在提高检测率的同时,降低了误检率,具有较好鲁棒性。
关键词: 人脸检测 SURF特征 FAST算法 Adaboost算法 人眼定位
For information in English, please click here
Face detection under rotation in image plane based on improved Speeded Up Robust Features
Abstract:To solve the problem of face detection under rotation in image plane, the paper proposes an improved SURF algorithm combined with Adaboost algorithm and fusion of human eyes to realize the plane rotation face detection. First, to improve the SURF algorithm, the FAST algorithm is used instead of the Hessian matrix to detect the feature points.The improved SURF algorithm and Adaboost algorithm are combined to train a set of SURF face common feature set. Then, the SURF feature of the image is extracted and the eye location is used to filter the invalid key points.Finally, the extracted features are matched with the common features of the face, and achieve the face detection.Experiment result express thatthe algorithm has great advanFace detection under rotation in image plane based on improved Speeded Up Robust Featurestages in the detection rate and false detection rate.It can improve the detection rate and reduce the false detection rate. The algorithm has high robustness.
Keywords: face detection SURF feature FAST algorithm Adaboost algorithm eye location
论文图表:
引用
No.4719408118414114****
同行评议
共计0人参与
勘误表
基于改进加速鲁棒特征的平面旋转人脸检测算法
评论
全部评论0/1000