一种改进的注射药液不溶异物检测算法

An Improved Algorithm for Detecting Insoluble Foreign Substances in Liquid Transfusion

注射药液生产过程中容易混入不溶异物,这些异物一旦进入人体会影响人体健康甚至威胁生命,因此在药液投入使用前检测出这些不溶异物至关重要。设计一套针对安瓿瓶药液内异物的视觉检测算法,利用序列图像提取药液中的不溶物体,然后进行面积判定及异物判别。对传统滤波与帧间差分过程进行改进,以消除序列图像中存在的大量背景噪声。实验结果表明,该改进算法具有检测速度快、精度高等特点,检测准确率达到96%,能够满足实际生产需求。

It is common to mix insoluble particles into liquid transfusion in the production process. Once these foreign particles enter the human body,it will affect human health and even life. Therefore,it is very important to detect these insoluble foreign substances before injection. In this paper,a set of visual detection algorithm for the foreign substances in the ampoule liquid was designed. The se-quence images were used to extract the insoluble particles in the liquid in order to identify them. Corresponding improvements have been made to the traditional filtering and inter-frame difference processes to eliminate the large amount of background noise present in the sequence images. Preliminary experimental results indicate that the algorithm has fast detection speed,high precision rate,and the detection accuracy rate reaches 96%,which can satisfy the requirement of the medicine production lines.