基于液面分割的安瓿内漂浮小目标检测算法

为了检测出安瓿瓶内的弱小漂浮目标,提出了一种基于液面边缘分割的检测算法。首先计算序列图像的前三阶Legendre正交矩值,选取合适的判断阈值参数来获取液面边缘;然后利用局部曲线拟合修复边缘断点得到完整的边缘;最后分割出液面区域并作序列图像的帧间差分运算,获取漂浮小目标的运动轨迹。实验结果表明:相对于传统的帧间差分法,该算法很好地去除了液面噪声,能够提取弱小漂浮目标的运动轨迹,满足在生产中对漂浮目标的提取和快速识别的要求。

Floating impurity inspection algorithm in ampoule based on liquid level segmentation

An inspection algorithm based on liquid level segmentation was developed to detect small floating targets in ampoules.The algorithm calculates the first three-order orthogonal Legendre moment of the image sequences and selects the appropriate threshold to determine the liquid edge.The complete edge is then obtained by local curve fitting.Finally,liquid level areas are segmented and the differences in image sequences are analyzed to obtain the trajectory of small floating targets.Tests show that this algorithm eliminates the noise generated by the liquid level better than traditional frame difference methods to efficiently extract the trajectories of small floating impurities for rapid recognition of floating particles during production.