基于轻量化卷积神经网络的隧道勘察地层岩性划分方法  

Stratum Lithology Division Method for Tunnel Survey Based on Lightweight Convolutional Neural Network

作　　者：丁海英[1] DING Haiying(College of Coal and Chemical Industry, Shaanxi Energy Institute, Xianyang 712000, China)

机构地区：[1]陕西能源职业技术学院煤炭与化工产业学院,陕西咸阳712000

出　　处：《微型电脑应用》2021年第10期190-193,共4页Microcomputer Applications

摘　　要：目前在隧道勘察过程中对地层岩性进行划分时,没有对地层图像进行预处理,导致划分结果准确率低、召回率低、F-measure低。提出基于轻量化卷积神经网络的隧道勘察地层岩性划分方法,采用灰度变化增强方法对地层图像进行增强处理,避免噪声对地层岩性划分产生干扰,通过OTSU算法对地层图像进行分割处理,提取目标地层区域,消除岩屑颗粒重叠现象或无关背景对提取地层岩性特征产生的干扰。采用数理统计分析方法提取预处理后的地层图像的纹理特征,构成纹理特征向量,在轻量化卷积神经网络中输入纹理特征向量,实现地层岩性的划分。实验结果表明,所提方法的准确率高、召回率高、F-measure高。Currently,in the classification of the stratum lithology in the tunnel survey process,the stratum image is not preprocessed,which results in low accuracy,low recall,and low F-measure.A lightweight convolutional neural network-based tunnel survey stratum lithology division method is proposed.The gray-scale change enhancement method is used to enhance the stratum image to avoid noise interference to the stratum lithology division.The stratum image is segmented by the OTSU algorithm.It extracts the target stratum area to eliminate the interference of the overlapping phenomenon of cuttings particles or irrelevant background on the extraction of lithological characteristics of the stratum.The mathematical statistical analysis method is used to extract the texture features of the preprocessed stratum image to form a texture feature vector.The texture feature vector is input into the lightweight convolutional neural network to realize the division of stratum lithology.Experimental results show that the proposed method has high accuracy,high recall and high F-measure.

关 键 词：轻量化卷积神经网络 隧道勘察 地层岩性划分 图像增强 OTSU算法 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TN911.73[自动化与计算机技术—计算机科学与技术]