摘要
针对输液可见异物检测问题,提出了一种基于图像序列的多个可见异物检测算法。首先利用归一化的互相关系数和自适应迭代求解阈值获得图像的二值图像,然后在传统的均值漂移(meanshift)算法的基础上引入带宽自适应核函数和自适应变步长因子,解决可见异物可能发生形状变化和算法收敛速度慢的问题。实验结果表明,基于该算法的检测方法可以快速且较准确地同时检测出输液中的多个异物。
Aiming at detecting problem of foreign substance in fluid infusion,a multi-target detecting algorithm based on sequence images was proposed.Firstly,binary image was obtained by using the normalized cross-correlation coefficient and self-adaptive iterative method.Then,on the basis of traditional Mean Shift algorithm,the bandwidth adaptive kernel function and the adaptive variable step-size factor were proposed to solve the problems of the changeable shape of particle and the slow convergence of the algorithm.The experiment results showed that detection methods based on the algorithm could quickly and more accurately detect several foreign substances in fluid infusion.
出处
《武汉理工大学学报（信息与管理工程版）》
CAS
2010年第4期565-568,共4页
Journal of Wuhan University of Technology：Information & Management Engineering
基金
教育部博士点专项基金资助项目(20070497105)
关键词
序列图像
异物检测
图像处理
归一化互相关系数
均值平移
sequence image
foreign substance detection
image processing
normalized cross correlation coefficient
mean shift
