基于小波变换和概率神经网络的焊接接头类型识别  

Classification of welding joints based on wavelet transform and probabilistic neural network

作　　者：王秀平[1,2] 白瑞林 刘子腾 陈晶[1,2] 

机构地区：[1]江南大学轻工过程先进控制教育部重点实验室,江苏无锡214122 [2]无锡科技职业学院,江苏无锡214028

出　　处：《光学技术》2015年第2期138-143,共6页Optical Technique

基　　金：江苏省高校优势学科建设工程资助项目(PAPD);江苏省产学研前瞻性联合研究项目(BY2012056);江苏省自然科学基金项目(BK20141115)

摘　　要：为了实现机器人自动焊接过程中快速、精确地提取焊缝特征信息,提出了一种基于小波变换和概率神经网络的焊接接头类型识别方法。先采用小波变换对由激光视觉传感器采集的焊接接头图像进行降噪和增强,对重构后的图像进行二值化,然后提取图像的特征信息,组成图像特征向量,最后构建概率神经网络分类器并进行测试。结合视觉传感器中激光器与摄像机的位置关系,最终识别出4种焊接接头。实验结果表明,所提出的方法特征提取简单,识别率高,并具有较好的实时性。A new method of welding joint recognition based on wavelet transform （WT） and probabilistic neural network （PNN） is proposed to extract features of weld seam rapidly and precisely in robotic welding. Images of welding joints captured by laser vision sensor are preprocessed firstly for noise reduction and enhancement, and the reconstructed images are converted to binary ones using appropriate thresholds. Then image features of binary images are further extracted and feature vectors are formed to input into a probabilistic neural network classifier for classification. Combined with the position relationship of laser and camera, four types of welding joints are eventually recognized. The experimental results show that the proposed method is simple, efficient and can get high recognition rate and good real-time performance.

关 键 词：光学测量 小波变换 概率神经网络 焊接接头 分类器 

分 类 号：TP242.6[自动化与计算机技术—检测技术与自动化装置]