问寻

摘 要：针对实验玻璃器皿高质量要求，提出一种基于图像处理的实验室玻璃器皿缺陷检测方法。为更直观检测到玻璃器皿的缺陷，以实验室常用的西林瓶为例，首先对瓶口检测区域进行定位，将瓶口与背景图像分割，并利用最大类间方差法对图像进行填充，从而得到瓶口图像的定位;将定位好的瓶口图像通过极坐标变换和动态阀值分割，实现瓶口图像特征的提取，最后运用模板匹配法实现对瓶口图像缺陷的检测。结果表明，本缺陷监测方法可很好实现毛口和破口的识别，且在不同瓶口半径下的缺陷检测平均准确率可达97.4%，具有较高的识别准确率。

关键词：实验用玻璃器皿;缺陷检测;模板匹配;准确率

中图分类号：TQ171 文献标识码：A 文章编号：1001-5922（2021）08-0054-04

Research on Defect Detection of Experimental Glassware

Wen Xun

（Xi an Medical College， Xi an 710100， China）

Abstract：Aiming at the high quality requirements of laboratory glassware， a defect detection method of experimental glassware based on image processing is proposed. In order to detect the defects of glassware more intuitively， taking the commonly used penicillin bottle in the laboratory as an example， firstly， the detection area of the bottle mouth is located， the bottle mouth and the background image are segmented， and the Otsu method is used to fill the image， so as to get the positioning of the bottle mouth image. The positioned bottle mouth image is segmented by polar coordinate transformation and dynamic threshold to realize the extraction of bottle mouth image features. Finally， the template matching method is used to detect the defects of the bottle mouth image. The results show that this defect monitoring method can well realize the recognition of rough mouth and break， and the average accuracy of defect detection under different bottle mouth radius can reach 97.4%， which has high recognition accuracy.

Key words：experimental glassware; defect detection; template matching; accuracy

眾所周知，实验室对玻璃器皿的质量要求很高。如玻璃器皿存在缺陷，可能导致实验事故。因此，做好玻璃器皿的缺陷检测，是提高实验安全的重要途径。现代玻璃器皿生产工艺的缺陷很多，如破口、裂纹以及黑点等。以典型的西林瓶为例，西林瓶是实验室比较常见的一种玻璃管制注射剂瓶。一旦出现裂纹，很容易造成药品的污染，进而危及患者安全[1]。因此，加强对西林瓶缺陷的检测，是当前思考和研究的重点。传统的西林瓶缺陷检测通常采用图像视觉技术，以提高缺陷识别的准确率确率。本研究则提出对西林瓶瓶口图像进行处理与识别，进而判别瓶口的缺陷问题。

1 瓶口检测区域定位