中文摘要：

针对布匹瑕疵种类多以及单个方法仅对特定类瑕疵有效的问题,提出一种新的基于小波多尺度积和数学形态学的布匹瑕疵检测算法。首先对输入图像进行二进小波变换后,低频近似子图经过数学形态学运算,得到良好的瑕疵形状特征,然后对高频子图使用小波多尺度积方法,可以抑制噪声的同时增强瑕疵的边缘线性特征,最后融合得到最终的检测结果。实验结果表明,该算法在虚警率和运算时间较低的同时,得到较高的检测率,综合性能优于经典的Gabor和小波变换算法。

英文摘要：

The number of fabric defect classes is large and a single method is only effective for specific types of defects.In order to detect defects more effectively,a novel fabric defect detection algorithm is presented based on combining wavelet multi-scale product and mathematical morphology.After the dyadic wavelet transform on the images,mathematical morphology operations are utilized to the low frequency sub-image to obtain defect shape features.Then,the wavelet multi-scale product methods are used to the high frequency sub-images to realize suppressing noise and enhance defect edge linear features.Finally,the results above are fused to get the final result.Experimental results show that the false alarm rate and computational time are low,and the detection rate is high.The comprehensive performance is superior to the Gabor and wavelet algorithms.