摘要
提出一种基于ART神经网络(及其改型)从地震剖面中识别子波的新方法.发挥神经网络在模式识别和优化计算两方面的优势,先以无教师学习算法从剖面中识别出子波,再通过TH神经网络作反褶积.实验结果表明,本文提出的方法有明显的效果,特别在地震剖面中存在一些相隔较远的反射层时有很好的效果.
A new deconvolution method based on ART neural network and its derivative form, which have advantages in pattern recognition and optimization computing, is presented. With the new method the wavelet was first picked out from a seismic profile using a learning algorithm without tuitors, then deconvolution was achieved through TH neural network. Experimental results showed that the new method has good performances, especially when there exist rather seperated reflecting layers in the profile.