为了准确提取与识别焊缝射线数字成像中焊接缺陷，本文提出了一种基于神经网络的模式识别算法。首先，分析了非线性模式分类的基本原理，通过人工神经网络实现对焊缝内存在的焊接缺陷进行分类；然后，采用缺陷的几何特征作为分类算法的输入数据，并应用神经网络关联标准理论评估鉴别能力，证明了特征提取的质量重要性优于数量；最后，将基于神经网络的主要非线性鉴别分量的识别算法应用于缺陷识别中，并通过大量实验分析与评价其分类性能。实验结果数据证明该算法在焊接缺陷模式识别方面具有较高的效率。%In order to extract and recognize welding defects in digital X-ray images, this paper proposes a neural network based on pattern recognition algorithm. Firstly, the fundamental of the nonlinear pattern classification has been analyzed. By means of artificial neural network, the classification of welding defects in welding lines has been realized. Later on, the geometric feature of the welding defect has been adopted for input data. The identification ability was evaluated by neural network association standard theory. It proved that quality was more important than quantity. At last, the neural network based on principal discrimination components has been applied to defect identification and satisfying result has been achieved. The experimental result proved this necognition has high efficiency.
展开▼
