基于支持向量机的磁粉探伤智能识别算法  

Intelligent Recognition Algorithm Based on Support Vector Machine Magnetic Particle Detection

作　　者：李远江[1] 张华[1] 王姮[1] 张静[1] 

机构地区：[1]西南科技大学特殊环境机器人技术四川省重点实验室,四川绵阳621010

出　　处：《西南科技大学学报》2015年第1期66-70,共5页Journal of Southwest University of Science and Technology

基　　金：四川省科技支撑计划项目(2013GZX0152);四川省科技厅科技支撑计划项目(2014RZ0049)

摘　　要：磁痕图像的自动分类方法是磁粉探伤智能化的关键技术之一。针对目前磁粉探伤自动识别漏识率和虚警率高的问题,提出采用支持向量机算法后验概率输出的判断方式对磁痕图像进行分类。对采集到的图像进行预处理和特征提取,然后使用支持向量机对特征数据处理得到后验概率,利用此后验概率得到分类信息将结果映射为有无裂纹与类别模糊区三类,实现智能识别。实验证明算法的有效识别率高,在漏检率和虚警率上取得了较好的平衡。The automatic classification method for magnetic images is one of the key technologies of intelligent magnetic particle inspection. Aiming at the problem of identifying both leakage recognition rate and false alarm rate by the magnetic particle detector automatically, a method using support vector machine algorithm outputs to posterior probabilities to classify the magnetic images is proposed. After pre -processing the collected ima- ges and extracting their feature accordingly, the support vector machine was used to process characteristic data to get a posteriori probability, which was then used to obtain the classified information of the following three categories : cracks, no cracks and a fuzzy area, and finally an intelligent recognition system was achieved. The experimental system was established to prove the effective recognition rate of the algorithm and a better balance between the leakage recognition rate and false alarm rate was obtained.

关 键 词：磁粉探伤 支持向量机 类别模糊区 漏检率 虚警率 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]