作　　者： ;

机构地区： 东莞职业技术学院

出　　处： 《计算机测量与控制》 2014年第6期1676-1679,共4页

摘　　要： 根据织物检测的实际情况需要,提出了基于差分盒算法的改进算法,在使用Brodatz纹理库样本的前提下,分别在盒子高度确定、盒子总数统计以及网格中盒子数量确定3个方面进行改进,在算法时间和精度两项上,对经典差分盒算法和改进算法进行了比较,对比得出了改进算法的时效性;还应用了改进差分盒算法对3种常见纹理的疵点织物进行了计算,以确认疵点织物,并验证了改进差分盒算法;实验通过检出率、误检率、漏检率和检测精度4个检测精度参数表明,改进的差分盒算法可以有效地区分疵点织物和正常织物,该方法具有很强的实用性。 For the need of detection of fabric defects, the paper presents an improved algorithm based on differential box--counting （ID- BC） method. The algorithm modifies the highness of the boxes, the method of determining the amount of boxes in grids and the statistics of calculating the total of all the boxes on the premise of using Brodatz texture library sample. On the two aspects of the time and accuracy, the paper compared the new algorithm with classical differential box--counting method （DBC） and obtained the efficiency of＇the improved algo rithm. At the same time, the paper calculated defect fabric of the three common texture employing improved differential box--counting meth- od to verify defect fabric. Through four parameters, i. e, detection rate, false drop rate, loss rate and detection precision, the results of ex periments demonstrate this new method is effective and can outperform the well--known differential boxing--counting （DBC） method.

关 键 词： 差分盒 纹理图像 机器视觉 分形维数

领　　域： [自动化与计算机技术] [自动化与计算机技术]