基于改进LeNet-5的油井井号识别方法
|
摘要
针对目前油井井号识别效率低,识别方法效果差的问题,本文采用基于深度学习的卷积神经网络(ConvolutionalNeural Network,CNN)方法进行识别。针对传统卷积神经网络LeNet-5结构应用在井号识别中存在的不足,本文通过转换模型中卷积核大小、增加高效降维层、使用混合激活函数,三方面进行改进。经过实验,改进后的LeNet-5模型与传统网络结构相比,减少训练时间、提高准确率,在油井井号识别上具有明显的优势。
In this paper, the Convolutional Neural Network(CNN) method based on depth learning is used to identificate oil well numbers because of the low efficiency in oil well numbers recognition and the poor effectiveness of identification methods.In view of the shortcomings of the traditional convolution neural network LeNet-5 architecture in the recognition of well numbers, the three aspects are improved in this paper. They are transforming the convolution kernel size, increasing the efficient dimension reduction layer and using the mixed activation function. Through experiments, the training time of the improved LeNet-5 model is reduced and accuracy is improved compared with traditional network structure. It is obviously better than the traditional network structure in the identification of oil well numbers.
In this paper, the Convolutional Neural Network(CNN) method based on depth learning is used to identificate oil well numbers because of the low efficiency in oil well numbers recognition and the poor effectiveness of identification methods.In view of the shortcomings of the traditional convolution neural network LeNet-5 architecture in the recognition of well numbers, the three aspects are improved in this paper. They are transforming the convolution kernel size, increasing the efficient dimension reduction layer and using the mixed activation function. Through experiments, the training time of the improved LeNet-5 model is reduced and accuracy is improved compared with traditional network structure. It is obviously better than the traditional network structure in the identification of oil well numbers.
引文
[1]刘睿强.深度学习算法在船舶电子海图识别中的应用[J].舰船科学技术,2017,(2x):79-81.
[2]程国建,刘丽婷.深度学习算法应用于岩石图像处理的可行性研究[J].软件导刊,2016,15(9):163-166.
[3]蔡娟,蔡坚勇,廖晓东等.基于卷积神经网络的手势识别初探[J].计算机系统应用,2015,24(4):113-117.
[4]龚丁禧,曹长荣.基于卷积神经网络的植物叶片分类[J].计算机与现代化,2014,(4):12-15.
[5]曹林林,李海涛,韩颜顺等.卷积神经网络在高分遥感影像分类中的应用[J].测绘科学,2016,41(9):170-175.
[6]LECUN,BOTTOU.Reading Checks with Multilayer Graph Transformer Networks[J].1997,(1):151-154.
[7]PAULIN M,MAIRAL J,DOUZE M,et al.Convol utional Patch Representations for Image Retrieval:An Unsupervised Approach[J].International Journal of Computer Vision,2016,121(1):1-20.
[8]涂春鸣,罗安,刘娟.无源滤波器的多目标优化设计[J].中国电机工程学报,2002,22(3):17-21.
[9]胡正平,何薇,王蒙等.多层次深度网络融合人脸识别算法[J].模式识别与人工智能,2017,30(5):448-455.
[10]王雪光,郭艳兵,齐占庆.激活函数对BP网络性能的影响及其仿真研究[J].自动化技术与应用,2002,21(4):15-17.
[11]SRIVASTAVA N,HINTON G,KRIZHEVSKYA,et al.Dropout:a simple way to prevent neural networks from overfitting[J].Journal of Machine Learning Research,2014,15(1):1929-1958.
[12]刘杰辉,范冬雨,田润良.基于ReLU激活函数的轧制力神经网络预报模型[J].锻压技术,2016,41(10):162-165.
[2]程国建,刘丽婷.深度学习算法应用于岩石图像处理的可行性研究[J].软件导刊,2016,15(9):163-166.
[3]蔡娟,蔡坚勇,廖晓东等.基于卷积神经网络的手势识别初探[J].计算机系统应用,2015,24(4):113-117.
[4]龚丁禧,曹长荣.基于卷积神经网络的植物叶片分类[J].计算机与现代化,2014,(4):12-15.
[5]曹林林,李海涛,韩颜顺等.卷积神经网络在高分遥感影像分类中的应用[J].测绘科学,2016,41(9):170-175.
[6]LECUN,BOTTOU.Reading Checks with Multilayer Graph Transformer Networks[J].1997,(1):151-154.
[7]PAULIN M,MAIRAL J,DOUZE M,et al.Convol utional Patch Representations for Image Retrieval:An Unsupervised Approach[J].International Journal of Computer Vision,2016,121(1):1-20.
[8]涂春鸣,罗安,刘娟.无源滤波器的多目标优化设计[J].中国电机工程学报,2002,22(3):17-21.
[9]胡正平,何薇,王蒙等.多层次深度网络融合人脸识别算法[J].模式识别与人工智能,2017,30(5):448-455.
[10]王雪光,郭艳兵,齐占庆.激活函数对BP网络性能的影响及其仿真研究[J].自动化技术与应用,2002,21(4):15-17.
[11]SRIVASTAVA N,HINTON G,KRIZHEVSKYA,et al.Dropout:a simple way to prevent neural networks from overfitting[J].Journal of Machine Learning Research,2014,15(1):1929-1958.
[12]刘杰辉,范冬雨,田润良.基于ReLU激活函数的轧制力神经网络预报模型[J].锻压技术,2016,41(10):162-165.