摘要
以表面信息为基础进行火花塞缺陷检测时,大多通过人眼观测获取表面信息,导致检测结果F1-score值较低。因此,提出应用机器视觉技术的火花塞缺陷自动检测方法。依托于机器视觉技术原理,布置工业摄像机与CMOS型图像传感器,在合适的光照条件下采集包含详细表面信息的火花塞图像。对采集图像进行灰度转换、中值滤波以及配准处理,得到有效处理后火花塞图像。通过改进HOG特征提取算法,抽取火花塞图像包含的关键特征信息,输入支持向量机分类器中,得到缺陷自动检测结果。实验结果表明:所提检测方法的F1-score值为0.97,与基于特征融合、基于决策树的方法相比,F1-score值分别提升16%、21%。
Based on the surface information,when the spark plug defect detection is carried out,most of the surface information is obtained through human eye observation,resulting in the low F1-score value of the detection result.Therefore,an automatic detection method of spark plug defects based on machine vision technology is proposed.Based on the principle of machine vision technology,industrial cameras and CMOS image sensors are arranged to collect spark plug images containing detailed surface information under appropriate lighting conditions.After gray-scale conversion,median filtering and registration of the collected image,the effectively processed spark plug image is obtained.By improving the hog feature extraction algorithm,the key feature information contained in the spark plug image is extracted and input into the support vector machine classifier to obtain the automatic defect detection results.The experimental results show that the F1-score value of the proposed detection method is 0.97.Compared with the method based on feature fusion and decision tree,the F1-score value is increased by 16%and 21%respectively.
作者
谭立志
TAN Lizhi(Hunan Automobile Engineering Professional College,Zhuzhou,Hunan,412001,China)
出处
《小型内燃机与车辆技术》
CAS
2023年第5期81-86,共6页
Small Internal Combustion Engine and Vehicle Technique
基金
湖南省自然科学基金项目“基于机器视觉的火花塞表面质量缺陷检测关键技术研究”,编号:2021JJ60055。
关键词
机器视觉技术
火花塞
表面缺陷
特征提取
图像配准
支持向量机
Machine vision technology
Spark plug
Surface defects
Feature extraction
Image registration
Support vector machine
