摘要
本文提出了一种全新的基于矩形拟合的特征算法,即在图像中,用颜色分割将人体的四肢和躯干分别用矩形进行表示,然后找出各个矩形区域的面积、中心点、长短轴等作为每个人体部位的特征,再将人体每个部位的这些特征与设定的阀值相比,判断是否为人体,从而实现人体检测。实验结果表明,提出的新方法能够运用于人体的检测并得到良好的检测结果。
This paper proposes a new feature algorithm based on rectangle fitting. The limbs and the trunk of human bodies are represented by using color segmentation in the image. The algorithm calculates the area,the center point,and the major and minor axis of each rectangular,and then compares these data with the threshold,so as to achieve the detection of human-body. The experimental results show that the new method can be applied to the detection of human body and can get a good result.
出处
《科技广场》
2016年第3期11-16,共6页
Science Mosaic
关键词
人体特征
颜色分割
矩形拟合
人体检测
Human Physical Characteristics
Color Segmentation
Rectangle Fitting
Human Detection
