应用视觉显著性的小提花织物疵点检测

为实现小提花织物的疵点检测，提出了一种基于视觉显著性的疵点定位与分割方法。针对小提花织物的花型具有周期性的特点，通过对织物图像进行快速傅里叶变换和形态滤波来抑制正常花纹的显著性，突出疵点区域的显著性，以获取图像的全局显著图；然后使用基于图论的视觉显著模型来计算图像的局部显著图，并对全局和局部显著图进行合并生成综合显著图；最后使用最大熵法对综合显著图进行分割，以得到疵点目标。实验结果表明，在对横裆、破洞、断头、打结和跳花等5 种瑕疵进行测试时，该方法的正确率高达93.5%，非常适合于对小提花织物进行疵点检测。

Defect detection for mini-jacquard fabric based on visual saliency

This paper proposed a new defect detection method for mini-jacquard fabric based on visual saliency. This method firstly analyzed the characteristic of the mini-jacquard and proposed to apply fast Fourier transform and mathematical morphological filter on the original image to keep down the saliency of the normal area, pop out the saliency of the defect area andobtain the global saliency map; and then, the local saliency map was obtained by using the graph-based visual saliency model; after that, the saliency map could be generated by combine the global and local saliency map; and finally,a maximum entropy method was implemented on the saliency map to separate the defected area. Experimental results show that the proposed method can effectively detect multiple kinds of defects of barre, hole, broken end, knots and netting,and the average success rate is 93.5%. It is an effective defect detection method.