基于高阶统计的放射图像自动焊接缺陷检测  

An Automatic Weld Defects Detection Method of Radiographic Images Based on High-order Statistic

作　　者：刘晓阳[1] 酆烽[2] 

机构地区：[1]济南职业学院电子工程系,山东济南250002 [2]山东钢铁集团济南钢铁股份有限公司,山东济南250101

出　　处：《铸造技术》2015年第9期2385-2389,共5页Foundry Technology

基　　金：山东省高等学校科研计划资助项目(J10LG63)

摘　　要：为了提高放射图像的焊接缺陷检测精度,引入了功率倒谱技术,并设计焊接区域定位机制,提出了基于高阶统计的放射图像自动焊接缺陷检测方法。首先,利用自适应直方图均衡技术增强放射图像;并使用均值维纳滤波器对增强后的图像进行滤波处理;然后通过设计焊接区域定位机制,从放射图像中确定焊接缺陷所在的区域;再引入功率倒谱技术,从高阶频谱(三阶谱)中提取放射图像的倒谱特征信息;并利用神经网络技术对提取信息进行特征匹配。仿真结果显示,与当前缺陷检测方法相比,本文方法的检测率要高于其他方法。In order to improve the detection precise of radiographic images of weld defects, an automatic weld defects detection method based on high-order statistic was proposed by introducing power cepstrum technology and designing the positioning mechanism of weld area. First of all, the adaptive histogram equalization technology was used for enhancing radiation images, and then the images were filtered by using the average wiener filter. After that, the areas of weld defects were determined from the radiographic images by the as-designed positioning mechanism of weld area. Then, cestrum features were extracted from the higher-order spectrums （trispectrum） by introducing power cepstrum technology. Finally, neural networks were used for matching the feature information. The simulated results show that the detection rate of the proposed method is higher than those of the other current-used radiography methods for defects detection.

关 键 词：高阶统计 功率倒谱 焊接缺陷检测 放射图像 倒谱特征 

分 类 号：TG441.7[金属学及工艺—焊接]