应用最优Gabor滤波器的经编织物疵点检测

Warp knit fabric defect detection method based on optimal Gabor filter

针对经编织物疵点自动检测问题,提出了一种新的基于最优Gabor滤波器的经编织物疵点检测方法。具体可分为学习阶段和检测阶段;在学习阶段,对于无疵点的经编织物图像构造可调制的二维Gabor滤波器,采用量子行为粒子群优化(QPSO)算法对Gabor滤波器的参数进行优化,得到与无疵点的织物图像纹理特征最匹配的Gabor滤波器参数;在检测阶段,由学习阶段得到的最佳参数构造Gabor滤波器,用该滤波器对待检测织物图像进行卷积处理,然后再对得到的卷积图像进行二值化处理,最终识别出待检测织物是否有疵点存在。结果表明,该方法的检测率可以达到96.67%,具有很好的稳定性和鲁棒性,适合应用于工业生产。

Focusingonautomaticimageinspectionofwarpknitfabricdefectsintextileindustry,anewmethodforwarpknitfabric defect detection based on an optimal Gabor filter is presented. The proposed method consists of two process： the training and the inspection process. In the training process, the parameters of the 2D-Gabor filter can be tuned by the quantum-behaved particle swarm optimization （QPSO） algorithm to match with the texture features of a defect-free template acquired in prior. In the inspection process, each sample fabric image under inspection is convoluted with the selected optimized Gabor filter. Then a simple thresholding scheme is applied to generate a binary segmented result. Experimental results show that the detection rate of the proposed method can reach 96.67%. It has good performance of stability and robustness, suitable for industrial production.