基于机器视觉的数码管显示单元缺陷检测系统

A Defect Detection System of Nixie Tube Display Unit Based on Machine Vision

介绍了通过机器视觉和OCR识别方法来判断测温仪上LED数码管的工作状态的缺陷检测系统,以提高产品的质量,减轻质检人员负担.系统主要采用机器视觉识别出的LED字符与系统发出的命令字相对照的方法实现缺陷检测.首先通过CRC-16算法生成各个仪器的命令字,然后通过RS485总线分别向仪器发送显示字符命令,通过红色滤光片采用机器视觉采集LED图像,并对图像进行二值法处理、“闭”运算、腐蚀运算.在OCR识别中使用扩大训练字符集及大范围识别参数调整的方法,解决了LED数码管在OCR识别过程中遇到的断码、小数点误判及字符变形、变小问题.系统在投运后运行可靠,稳定,OCR识别误码率小于万分之五.

In this paper,a defect detection system is introduced to judge the working state of LED digital tubes by using of thermometers through machine vision and OCR identification methods,so as to improve product quality and reduce the burden of quality inspectors.The system mainly adopts the method of comparing the LED characters recognized by machine vision with the command words issued by the system to realize defect detection.Firstly,the command word of each instrument is generated by CRC-16 algorithm.Then,the display character command is sent to the instrument through RS485 bus,and the LED image is collected with machine vision through red filter,and binary processing,"close"operation and corrosion operation on the image are performed.In OCR recognition,the method of expanding the training character set and adjusting the recognition parameters in a large range is used to solve the problems of broken code,misjudgment of decimal point,character deformation and reduction encountered by LED nixie tubes in OCR recognition.The system operates reliably and stably after being put into operation,and the OCR identification error rate is less than 5/10000.