透明液体药剂中可见异物跟踪算法  

Tracking algorithm of visible particles in transparent liquid pharmaceutical

作　　者：许雪梅[1] 李丽娴[1] 张键洋[1] 倪兰[1] 黄征宇[1] 曹建[1] 

机构地区：[1]中南大学物理与电子学院,湖南长沙410083

出　　处：《浙江大学学报（工学版）》2012年第10期1822-1830,共9页Journal of Zhejiang University：Engineering Science

基　　金：国家自然科学基金资助项目(61172047;61071025)

摘　　要：为了实现液体药剂生产过程中的自动化质量检测,提高质检的速度和准确性,在智能视觉的基础上对透明液体药剂中可见异物的跟踪算法进行研究.采用中值滤波器对输液图像进行去噪,应用二次差分与能量累积算法对液体药剂中的可见异物进行检测,分别运用漫水填充法和实时运动模板提取异物目标的形态特征和运动特征;在分析异物的运动轨迹基础上,建立基于运动特征的区域匹配跟踪模型,给出搜索窗口的定位与尺寸大小选择原则,根据速度大小和运动方向的差异采取不同的跟踪策略,解决了异物目标在消失和重叠情况下的跟踪难题.实验结果表明,改进的区域匹配跟踪算法的错跟率和漏跟率分别为0%和2.5%,跟踪的正确率可以达到97.5%,且具有较高的实时性,能够满足实际生产的需要.The tracking algorithm of visible particles in transparent liquid pharmaceutical was analyzed based on intelligent vision technology in order to realize the automatic quality testing in the production of liquid pharmaceutical and improve the speed and accuracy of quality inspection. The median filter was adopted to denoise the corresponding images. The second difference with the energy accumulation algorithm was used to detect particles. Then the flood fill method and real-time motion template were applied respectively to extract the morphological characteristics and the direction of movement about particles. The region-based tracking model was established based on the analysis of particles＇ moving tracking. The selection principle of size and location of the search window was given. Different tracking strategy was adopted according to the differences of speed and direction in order to solve the tracking difficulties when particles disappear or overlap. Results show that the wrong-tracking rate and the missing- tracking rate of the improved regional matching algorithm is 0% and 2.5% respectively, and the tracking accuracy can reach 97. 5%. The improved algorithm has a better real-time quality and the needs of real-time production can be satisfied.

关 键 词：异物检测 特征提取 区域匹配 目标跟踪 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]