基于Camshift跟踪算法与SVM的大输液杂质检测识别方法研究

针对我国医药生产检测包装线上大输液杂质智能检测技术问题,提出了一种利用实时视频图像处理技术检测识别大输液杂质的方法.该方法对连续多帧被旋转的大输液瓶图像运用差分图像运动分析方法提取目标杂质;运用图像处理技术去除气泡噪声,准确分割目标杂质,采用Camshift跟踪算法连续跟踪几帧运动杂质以确保检测准确率;根据Camshift跟踪算法提取出的杂质运动和几何特征,应用SVM(Support Vector Machine)准确识别杂质类型.实验结果表明,该方法检测识别直径大于等于4个像素的杂质的检测识别率平均可达到95.4%,检测识别速率平均可达到581 ms/瓶.

Research on the Detection and the Recognition System of Impurity in Transfusion Solution Based on Camshift Algorithm and SVM

Considering that the intelligent detection and recognition of impurity in transfusion solution is a critically needed technology in pharmaceutical device production, a method for impurity detection and recognition using video image processing technology was proposed. Firstly, the impurity objects in several continuous rotated transfusion solution bottle frames were extracted cursorily by subtracting between two frames. Then, some image processing technologies were utilized to eliminate bubble noises and extract impurity objects accurately. And then, the Camshift tracing algorithm was adopted to trace the impurity to ensure the correct detection. Finally, Support Vector Machine was used to recognize the type of impurities with the extracted motion and geometrical features of impurities. Experiment results have shown that the mean detection and recognition rate is 95.4% for impurities equal and larger than four pixels in diameter. Mean detection and recognition time is 581ms for one transfusion solution bottle.