摘要
针对热敏电阻封装质量检测问题,设计了基于阈值分割及轮廓矩识别的自动检测方法。该方法首先对封装热敏电阻进行灰度化处理,再利用最大类间方差法对图像进行阈值分割,进一步通过形态学运算得到不含噪声的目标二值图像,并采用Canny算子提取到封装热敏电阻的连续轮廓,然后根据轮廓矩理论提取了封装热敏电阻边缘图像的前三阶轮廓矩,最后得到与模板的欧氏距离,通过阈值比对实现了封装质量识别。实验验证表明该方法对几种封装不合格产品的固定样本检测准确率都在90%以上,对随机样本的检测准确率接近人工检测的准确率,且单根检测耗时220 ms远远小于单根平均5 s的人工检测耗时。
An automatic detection method based designed to solve the problem of thermistor pac on threshold segmentation and contour moment recognition is kaging quality identification. First, the package thermistor is converted to gray scale and the threshold is segmented by the Otsu method, after that, the target binary image with noise is obtained by morphological operation and continuous contour of the package thermistor is extracted by means of the Canny operator. Then the first three-order contour moments of the edge image of the package thermistor are calculated according to the contour moment theory. Package quality identification of thermistor is finally realized through the calculation of Euclidean distance from the template and threshold determination. It is shown that fixed sample detection accuracy is more than 90% for several packages of substandard products, the detection accuracy of random samples is close to the accuracy of manual detection, and single detection time takes 220 ms, which is far less than the single average of 5 s of manual detection time.
作者
刘碧俊
LIU Bi-jun(School of Mechanical and Electrical Engineering, Jiangsu Food ＆ Pharmaeeutieal Science College, Huai＇ an 223003, Chin)
出处
《测控技术》
CSCD
2017年第11期45-49,共5页
Measurement & Control Technology
基金
淮安市科技支撑计划(工业)项目(HAG2013029)
关键词
热敏电阻
封装质量
机器视觉
阈值分割
轮廓矩
thermistor
package quality
machine vision
threshold segmentation
contour moment
