摘要
针对数码印花时喷头堵塞和步进偏差等原因造成的周期性缺陷问题,提出一种融合小波分解的局部二值模式方差(LBPV)数码印花缺陷检测方法。对印花缺陷图像进行多尺度小波分解,选定高频细节子图进行图像融合,有效抑制印花图像背景纹理;计算融合后子图LBPV特征来降低纹理特征的维度;对特征图在水平方向上投影并统计出特定LBPV值的像素点个数。根据投影曲线的峰值,实现对数码印花缺陷的检测与定位。实验结果表明:该算法可精确检测印花缺陷,并实现缺陷位置的可视化识别。
Aiming at solving the periodic defects of nozzle clogging and step error in printing process,the method of digital printing detect detection based on wavelet decomposition and local binary pattern variance(LBPV)feature is proposed.Firstly,the printing defect image is decomposed by wavelet transform and fused with specific high-frequency detail sub-images to effectively suppress the background texture of the printing image.Then,the local binary pattern variance feature of the fused sub-image is calculated to reduce the dimension of the low-texture feature.Finally,the feature map is projected in horizontal direction and the number of pixels of specific LBPV value is calculated.Furthermore,the digital printing defects was detected and located by the peak value of the projection curve.In our research,the proposed algorithm can detect the printing and realize the visual recognition of defect locations accurately.
作者
苏泽斌
高敏
李鹏飞
景军锋
张缓缓
SU Zebin;GAO Min;LI Pengfei;JING Junfeng;ZHANG Huanhuan(College of Electronic Information,Xi’an Polytechnic University,Xi’an 710048,China)
出处
《传感器与微系统》
CSCD
2020年第9期126-128,131,共4页
Transducer and Microsystem Technologies
基金
陕西省教育厅科研计划资助项目(18JK0339)
陕西省高校科协青年人才托举计划资助项目(20180115)。