应用ART2人工神经网络自组织形成焊缝图象中的典型空间模式

应用ＡＲＴ２人工神经网络算法，使采集到的焊缝横截面方向上的灰度分布数据自组织形成若干种空间模式，并把它们作为典型空间模式存储在ＡＲＴ２人工神经网络的ＬＴＭ中。对实时采样到的灰度分布进行空间模式匹配程度检验，根据模式分布情况确定出焊缝位置。文中对梯度法检测结果进行了分析和比较，结果表明基于ＡＲＴ２人工神经网络的焊缝位置检测方法具有更强的噪声抑制能力，因而检测结果更准确、可靠。

The Self-Organization of Typical Space Patterns in Welding Seam Tracking Image by Using ART2 Artificial Neural Networks

The alogrithm based on the ART2 aftitical neural networks is used to self-organize the grey level distribution data in the transversal direction of a welding seam and forms several typical space patterns and stores them in the LTM (long time memory)of the ART2 artificial neural networds.Then the grey level distribution data of a welding image obtained in real time is matched with these typical space patterns.The welding seam position can be determined accoring to the situation of pattern distributions.It is shown that this method can inhibit strong noises and work correctly in a very nolsy circumstance.