欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
针对经典形状上下文算法对物体关节相对位置变化敏感的缺点,提出一种基于剪影局部形状填充率的物体识别算法.该算法以物体不同的轮廓控制点为圆心,计算不同半径下物体剪影像素所占总像素的比例,即为控制点的局部形状填充率;将不同控制点、不同半径长度所计算的形状填充率数值构成一个特征矩阵,该矩阵反映了物体整个剪影的统计特性.通过不同数据库的实验结果表明,文中算法对物体的细节有很强的描述能力,对物体关节的相对位置不敏感,并且受剪影轮廓控制点数量影响小.
英文摘要：
The shape context algorithm is sensitive to the variance of the articulation relative position.A shape recognition method is proposed in this paper by using the information of the local shape filling rate of object silhouettes.The ratio of the shape silhouette pixels over the total pixels is defined as the local shape filling rate and it takes the landmark point as the center of circle having different radius.The shape filling rates of all the landmark points with different radius constitute a characteristic matrix that reflects the object entire statistical property.The new approach is tested on a variety of shape databases.The novel method has strong power to describe the object details.It is insensitive to articulation and the quantity of landmark points.
同期刊论文项目
同项目期刊论文
期刊信息
位置：