应用上下文视觉显著性的色织物疵点检测

为实现色织物疵点的有效检测,提出一种应用上下文视觉显著性的疵点检测方法。根据上下文视觉显著性的原则,将织物图像分为大小相同的图像块;然后针对每个图像块,选取K个与其最相似的图像块计算与该图像块的差异值之和,用该差异值之和表示该图像块中心像素的显著性;从而生成一幅视觉显著性图;最后对显著性图进行阈值分割,得到色织物疵点的检测结果。为验证该算法的有效性,将带有纬缩、破洞和跳花等区域性疵点的素色、条纹和格子色织物图像作为样本进行检测。结果表明:该方法可较好地抑制不同种类织物的纹理背景,突出疵点区域,实现疵点的有效检测,该方法在色织物疵点检测上具有一定的可行性。

Yarn-dyed fabric defect detection based on context visual saliency

In order to facilitate the effective detection of yarn-dyed fabric defects, a defect detecting method based on context visual saliency was proposed. Using this method, the fabric image was firstly divided into image patches of the same size according to the principle of context visual saliency. Following that, for every image patch, a number (K) of image patches, most similar to the concerned image patch were selected, and the sum of the differences among the K image patches and the image patch of concern were calculated. The calculated sum of the differences was then used to represent the saliency of center pixel of the image patches, thereby generating a visual saliency map. Finally, the threshold of the saliency map was segmented to obtain the detection result of the yarn-dyed fabric defect. In order to verify the validity of the algorithm, the yarn-dyed fabric regional defect image samples with looped weft, holes and netting of color dots, color stripes and color checks were detected. The experimental results show that the proposed algorithm can suppress the texture background and highlight the defect area of different types of fabrics and achieve the effective detection of fabric defects, which indicates the effectiveness of the method for detecting defects in yarn-dyed fabrics.