CAJ | 학술논문

为实现快速和有效的织物瑕疵自动检测,提出了一种基于时间序列而不是图像的功率谱纹理分析方法.依据Burg auto regressive(AR)算法估计得到谱数据,从中提取能够反映纹理周期和取向等特点的特征,并首次采用支持向量数据描述模型来检测织物瑕疵纹理.对包含多种疵点的若干织物样本的检测结果表明,依照本文所述方案能够在保持较低的误譬率前提下达到较高的疵点栓出率,证明了所述方案的可行性.
위실현쾌속화유효적직물하자자동검측,제출료일충기우시간서렬이불시도상적공솔보문리분석방법.의거Burg auto regressive(AR)산법고계득도보수거,종중제취능구반영문리주기화취향등특점적특정,병수차채용지지향량수거묘술모형래검측직물하자문리.대포함다충자점적약간직물양본적검측결과표명,의조본문소술방안능구재보지교저적오비솔전제하체도교고적자점전출솔,증명료소술방안적가행성.
For realizing the fast and effective detection of fabric defects,this paper proposes a novel texture analysis approach based on the power spectral density of one-dimensional time series rather than on the two-dimensional image.By using Burg auto-regressive(AR)algorithm to estimate the spectral density,feature characterizing periodicity and the orientation of the fabric texture are extracted.The support vector data description model is firstly used to detect fabric defect textural.Experimental results indicate that the low false alarm rate and the low missing rate can be simultaneously obtained,thus providing the effectiveness of the proposed approach.