摘要
在石油测井工程图纸的曲线矢量化过程中,由于背景网格与曲线、曲线与曲线存在大量交织,致使曲线跟踪中断,需要人工判断走向,难于实现自动跟踪。提出一种基于离散型Hopfield神经网络(DHNN)的测井曲线交叉线识别方法。这个方法先设置8个标准方向样本,对网络进行训练。在曲线跟踪的过程中,当遇到交叉点,就进入交叉线识别,通过训练好的网络进行分支匹配,再结合宽度匹配正确判断曲线走向。理论研究和实验分析表明,采用该方法提高了交叉线识别的正确率,抗干扰效果较好。
In the process of vector quantization of well logging curve, background grids and other forms of curve interference discontinue curve tracking and hardly to achieve automatic tracking, so manual direction judgment is needed. This article proposes a methodfor well log curves intersection recognition using Discrete Hopfield Neural Network （ DHNN）. This method presets 8 standard direction samples for network training. During curve tracking, entering recognition status if comes across intersection, the algorithm makes accurate prediction of curve direction through branch match of well-trained Hopfield network, in combination with width match. Theoretical analysis and experiment demonstrates that this method improves the precision of intersection recognition and has a good result of anti-interference.
出处
《计算机应用》
CSCD
北大核心
2008年第8期2036-2039,共4页
journal of Computer Applications
基金
四川省科技攻关资助项目(05GG021-026-03)
