摘要
轮廓矩特征能够有效描述目标的形状。与基于区域的矩特征不同,轮廓矩扫描目标周长像素点时,时间和空间复杂度低,但轮廓矩公式的推导过程复杂。提出用参数方程描述轮廓曲线段的方法,基于格林定理仅利用一次二项式定理推导出轮廓矩的一般表达式。以舰船投影图像为例计算了常用低阶轮廓矩,实验表明,与区域矩特征相比运行速度快,且质心误差不超过2%。
Contour moments can effectively describe the objects＇ shape. Different from global moments, the contour moments scan all the pixels on the perimeter of object with a lower time and space complexity, but the process of derivation is very complicated. We tried to simplify the procedure by describing the contour curve section with parameter equation. Based on the Green theorem, binomial theorem was used only once while calculating the contour moments. Taking the ship projection images as examples, general expressions of contour moments were calculated. The experiments showed that it operated faster than area moments, and the centroid error was less than 2%.
出处
《电光与控制》
北大核心
2013年第1期1-4,共4页
Electronics Optics & Control
基金
国家“八六三”课题创新基金项目资助(2010AAJ133)
关键词
目标识别
矩特征
轮廓
格林定理
object recognition
moments feature
contour
Green theorem
