基于特征融合与决策树级联结构的多姿态人脸检测  

Multi-pose face detection based on feature fusion and decision tree cascade structure

作　　者：郭耸[1] 顾国昌[1] 蔡则苏[2] 刘海波[1] 沈晶[1] 

机构地区：[1]哈尔滨工程大学计算机科学与技术学院,哈尔滨150001 [2]哈尔滨工业大学计算机科学与技术学院,哈尔滨150001

出　　处：《沈阳工业大学学报》2012年第2期203-208,共6页Journal of Shenyang University of Technology

基　　金：国家"863"计划资助项目(2006AA04Z259);中央高校基本科研业务费专项基金资助项目(HEUCF100606;HEUCF100604;HEUCFZ1010);国家教育部博士点专项基金资助项目(20092304120013)

摘　　要：为了提高复杂模式下多姿态人脸检测的速度与性能,提出了一种基于特征融合与决策树级联结构相结合的多姿态人脸检测方法.该方法给出了基于形态学梯度的边缘方位场特征,并提出了基于Haar-like特征与边缘方位场特征相融合的AdaboostSVM算法.通过对决策树级联结构进行改进,将特征融合的AdaboostSVM算法与改进的决策树级联结构相结合进行多姿态人脸检测.实验结果表明,该方法能明显改善复杂模式下多姿态人脸检测的速度与性能.In order to improve the speed and performance of multi-pose face detection in complex patterns, a multi-pose face detection method based on feature fusion and decision tree cascade structure was proposed. The edge-orientation field features based on morphological gradient were given in the proposed method, and a AdaboostSVM algorithm based on the fusion of both Haar-like and edge-orientation field features was proposed. With the improvement of the decision tree cascade structure, the AdaboostSVM algorithm with the feature fusion was combined with the improved decision tree cascade structure to carry out the multi-pose face detection. The experimental results show that the proposed method can improve the speed and performance of the multi-pose face detection in complex patterns obviously.

关 键 词：计算机视觉 人脸检测 多姿态人脸 特征融合 边缘方位场特征 AdaboostSVM算法 决策树级联结构 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]