纱线质量光电信号的消噪处理和异常检测  

Signal denoising and anomaly detection in an optical yarn quality measurement system

作　　者：盛国俊[1] 董永贵[1] 

机构地区：[1]清华大学精密仪器与机械学系,精密测试技术及仪器国家重点实验室,北京100084

出　　处：《清华大学学报（自然科学版）》2010年第2期229-231,共3页Journal of Tsinghua University(Science and Technology)

基　　金：国家自然科学基金资助项目(60672012)

摘　　要：构建了一种在生产过程中检测纱线质量的光电传感器,用于检测纱线的纱疵和颜色异纤等信息。针对纱线时域信号信噪比低、疵点难以辨析的问题,采用K阶矩滤波和最小熵滤波两种方法进行消噪处理及异常检测。对纱线质量光电测量系统的实测信号处理结果表明:两种方法均可有效实现信号的消噪,从而提取出有效信息。K阶矩滤波器具有运算量小、对纱线中异常信号提取能力强的特点。最小熵滤波器则能够通过参数的调整,使处理后的纱线信号保留更多的局部特征信息。An optoelectronic sensor was constructed to evaluate yarn quality during production. The system can detect abnormal signals that correspond to yarn defects, such as neps and foreign fibers with different colors. However, the time domain signal has a low signal to noise ratio; and therefore, the system cannot effectively discriminate between defects. Two methods, a K-th moment filter and a minimum entropy filter, were developed for signal denoising and anomaly detection. Analysis of experimental signal~ acquired by the optical yarn quality measurement system indicates that the two methods are both effective for signal denoising so that the defects can be extracted easily. The K-th moment filter has the advantage of lower computational complexity and is more effective at extracting yarn faults, while the minimum entropy filter can retain more tiny local features in the processed signals by properly adjusting the parameters.

关 键 词：光电传感器 纱线信号 K阶矩 最小熵 异常检测 

分 类 号：TP212.9[自动化与计算机技术—检测技术与自动化装置]