盒维数特征是检测织物疵点的一个有力工具,但是直接在二维图像基础上提取盒维数特征涉及的计算复杂度较大,因此,讨论了以一维时间序列为基础的简化盒维数特征向量取代原特征向量的可行性.分别就若干种常见疵点对上述两种特征向量在疵点检测的准确性和运算复杂度方面进行了较为细致的试验对比研究.研究结果表明,简化盒维数特征向量与原特征向量相比,在多数情况下具有相当甚至更佳的检测准确率,同时运算时间大幅度减少.%Box-dimension feature is a powerful tool for fabric defect detection, but large computational complexity is usually involved in the extraction of box-dimension features based on the original two-dimensional image. Therefore, a simplified box-dimension feature vector is discussed to replace the original feature vector. A comparative study in terms of defect detection accuracy and computational complexity is made between the original feature vector and the simplified one on the basis of several common types of defect. Experimental results indicate that the simplified feature vector is similar or even better than the original one as far as accuracy is concerned, but much more economic in the computational time.
展开▼
