基于时空运动信息的药液异物自适应分割算法  

An adaptive segmentation algorithm for impurity in medicinal liquid based on spatial-temporal moving information

作　　者：段中兴[1] 郭媛[1] 

机构地区：[1]西安建筑科技大学信息与控制工程学院,陕西西安710055

出　　处：《光电子．激光》2009年第6期779-783,共5页Journal of Optoelectronics·Laser

基　　金：国家自然科学基金资助项目(50575168);陕西省教育厅自然科学专项基金资助项目(07JK281)

摘　　要：为了能够从药液视频序列图像中准确地提取出运动异物对象,提出一种基于帧差图时空运动信息的药液异物模糊自适应阈值分割算法。首先通过4帧序列图像的隔帧差分得到2幅差分图并分别划分为5×5的图像块;然后计算对应图像分块的4阶矩以实现异物的运动信息提取;最后采用自适应阈值依据帧差图对应图像块的4阶矩之差实现运动异物的分割,并经过形态学处理去除掉噪声和空洞。为使阈值能够跟随图像块灰度变化,阈值的调整采用模糊推理依据对应块灰度均值差和方差变化自适应实现。实验及实际测试结果表明,本文所提算法较好地满足了低对比度和局部光照变化的药液异物实时检测要求,是一种实用有效的图像分割方法。In order to extract the moving video object accurately, a novel fuzzy adaptive threshold segmentation algorithm of moving impurity in medicinal liquid based on spatial-temporal moving information is proposed. Firstly, two discontinuous frame difference images obtained from 4 consecutive frames are partitioned into 5× 5 small blocks respectively. Then, the fourth-order statistics of the corresponding block is computed to extract the moving information of impurity. Lastly,an adaptive threshold is used to seg- ment the objects from background based on the difference value of high order statistics of corresponding blocks,and a morphological filter is used to remove the noises and holes. To adapt to the change of the block grey values, a fuzzy inference method is adopted to adjust the threshold adaptively according to the difference value of variance and grey value mean of blocks. The experimental and factual testing results show that the proposed algorithm can meet the demands of real time detecting of moving impurity in medicinal liquid,and it is a practicable and effective image segmentation method.

关 键 词：药液异物 视频对象提取 高阶统计 模糊推理 自适应阈值 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]