基于HOG特征的织物瑕疵检测算法

传统织物瑕疵检测依赖人工,成本高且效率低.提出基于机器视觉和图像处理算法的织物瑕疵检测方法,采用方向梯度直方图描述织物纹理特征,通过计算局部区域的方向梯度直方图特征与无瑕疵织物图像特征的相似性来识别瑕疵,在织物图像数据集上的待测结果表明,此算法能够有效地提取出织物的瑕疵.

Fabric defect detection algorithm based on HOG features

With the shortcomings like artificial dependence,high costs and low efficiency of traditional fabric defect detection,a new detection algorithm which was based on HOG (Histogram of Oriented Gradient) feature is proposed in this paper.To improve several shortcomings of traditional methods of defect detection (e.g.high computational complexity,low defect detection accuracy),the texture features of certain fabric are constituted by computing and counting the HOG of the region of local image.Experimental results on the fabric image data sets show that the algorithm can accurately detect defects.