基于支撑矢量机的织物疵点识别算法

为了使用机器对织物疵点进行有效地检测和分类,提出了基于直方图统计和支撑矢量机的织物疵点识别算法。该算法运用直方图统计的方法,由概率统计生成直方波形,基于波形特征参数对比能准确定位织物纹理结构的异常位置,正确识别织物疵点,并将其作为支撑矢量机的输入参数,用于训练特征样本集,以获得支撑矢量,对待识对象进行识别,得到识别结果,在识别结果中寻找最优匹配,将待识图像归入最匹配类中。实验结果表明,该算法用于织物疵点检测是可行、有效的,可得到满意的识别结果。

Fabric defect detection based on support vector machine

For effective detection and classification of numerous kinds of fabric defect using a machine, an algorithm for fabric defect detection based on dimensional histogram statistic and support vector machine is proposed, which uses dimensional histogram statistic method to generate dimensional histogram waveform and precisely locates the abnormal site on the fabric texture by comparing the parameters of the waveform characteristics and correctly identifies the defect of the fabric. The parameters are input into the support vector machine and used to train the samples with typical characteristics for obtaining support vectors. The target fabric is detected for its defects and the results are compared to find out the optimal match. The images of the target fabric are put into the best matching group. The experiments show this method is feasible and effective in fabric defect detection and satisfactory identification result can be achieved.