基于改进型PCNN 的智能灯检机研究[J].docVIP

基于改进型PCNN的智能灯检机研究葛继，王耀南，张辉，周博文 （湖南大学电气与信息工程学院长沙410082） 设计了一种基于改进型脉冲耦合神经网络（pulse coupled neural networks, PCNN）的全自动智能灯检机，以实现药液中可见异物在线高速、高精度检测。为了从采集的复杂序列图像中提取出微小的运动目标，本文设计的机械装置能保证智能灯检机按照“旋转-急停-跟踪拍摄”这一检测流程，利用序列图像在时空上的连续性，克服大量背景干扰。由于存在连续图像中物体位置变化不大而导致漏检的可能性，相机采用外部等间隔时间触发拍摄模式。最后应用所述改进型脉冲耦合神经网络分割图像，并根据目标运动轨迹的连续性和平滑特点判断异物的存在性。实验表明，该智能灯检机能够有效检测出药液中的可见异物，检测速度、精度、误漏检率均能很好的满足医药生产线的需求。智能灯检机；改进型PCNN；异物检测；图像处理TP391　　文献标识码：A　　国家标准学科分类代码：510.8040 Study of intelligent inspection machine based on modified pulsecoupled neural network Ge Ji, Wang Yaonan, Zhang Hui, Zhou Bowen （Department of Control Science and Engineering ,Hunan University, Changsha 410082, China） AbstractA biologically inspired spiking neural network model, the modified pulse coupled neural network (PCNN), has been applied in a real-time computer vision system (automatic intelligent inspection machine) to detect foreign substances in glass bottles filled with transfusion solution with high speed and accuracy. To extract small objects in complicated images, the designed mechanism guarantees the intelligent inspection machine follow the detection procedures - “revolving-abruptly stopping-video tracking”. Lots of background disturbances can be overcome using the spatiotemporal continuity of the image sequence. The CCD cameras work in outside-trigger mode to avoid the possibility of omission derived from the trivial position changes in the images. The proposed modified PCNN is applied to detect the existence of foreign substances according to continuity and smoothness properties of the object moving trace. Experimental results indicate preliminarily that the inspection machine can detect the visual foreign substances effectively; and the detecting speed, accuracy and false drop rate also satisfy the requirement of the medicine production assembly. Key words：intelligent inspection machine; modified pulse coupled neural network; foreign substance detection; image processing 1　引　　言 澄明度是反应注射剂等药液质量的一项重要指标，而且与临床治疗有着密切的关系