何扬名, 杜建强, 肖贤波. 基于区域深度特征的人头检测方法[J]. 微电子学与计算机, 2013, 30(11): 39-42.
引用本文: 何扬名, 杜建强, 肖贤波. 基于区域深度特征的人头检测方法[J]. 微电子学与计算机, 2013, 30(11): 39-42.
HE Yang-ming, DU Jian-qiang, XIAO Xian-bo. Detecting Human Head by Depth Characteristics of Regions[J]. Microelectronics & Computer, 2013, 30(11): 39-42.
Citation: HE Yang-ming, DU Jian-qiang, XIAO Xian-bo. Detecting Human Head by Depth Characteristics of Regions[J]. Microelectronics & Computer, 2013, 30(11): 39-42.

基于区域深度特征的人头检测方法

Detecting Human Head by Depth Characteristics of Regions

  • 摘要: 根据摄像机自上而下拍摄检测人头的特点,提出了以目标及其周围像素为候选匹配区域进行匹配来计算视差的人头检测方法。该方法首先用左视图检测所有可能的人头目标,然后选取目标及其周围像素作为匹配的整体,根据一些匹配的约束条件,用其来和右视图进行匹配,算出以目标及其周围像素为整体的视差。将目标及其周围像素作为整体,克服了人头区域纹理比较弱额缺点,使得匹配的准确率有所提高,从而提高了视差计算的准确度。

     

    Abstract: According to the features of cameras shooting from up to down,a new method of detecting human head is put forward,which uses the pixels of the target and pixels around the target.First,detect all possible targets in the left image.Then chose the pixels of every target and the pixels around the target as an integrated object,and use them to match the right image according to some constraint conditions.Thus the disparity of the integrated object can be gotten.The method integrating the pixels of target and around the target overcomes the shortcoming of weak texture of human head, which would increase the accuracy of matching. So the accuracy of disparity can be increased,too.According to the disparity,some fake target can be excluded.And the accuracy of recognizing human head would be increased.

     

/

返回文章
返回