古生物化石智能化识别方法及应用

Method and Application of the Intelligent Fossil Identification

对古生物化石的准确识别,不仅在古生物学研究中非常重要,而且是油气勘探的重要环节。古生物化石既可以厘定地层年代和沉积环境,也可以判断烃源岩沉积期的有机质富集条件。该研究以种一级Globoquadrina dehiscens、Glo⁃borotalia fohsi、Globorotalia menardii三类样本为例,提出一项古生物化石智能化识别方法,该方法基于图像处理方法和深度学习技术,通过图像形态学定位,图像背景处理,图像边缘扩充,图像旋转,图像加噪等一系列方法将样本从384个扩充至12288个,再使用基于卷积神经网络的VGG16架构对样本进行分类,其分类准确率可达92%以上。该项识别方法在油气勘探领域具有较强的实用价值和推广前景。

The accurate identification of paleontological fossils is not only important in paleontological research,but also an important part of oil and gas exploration.Paleontological fossils can determine the stratigraphic age and sedimentary environment,as well as the organic matter enrichment conditions during the deposition of source rocks.In this research,a set of Globoquadrina dehiscens,Globorotalia fohsi,and Globorotalia menardii samples in the species identification is used as an example to propose a kind of paleo-fossil intelligent identification methods based on image processing methods and deep learning techniques.The series of methods such as image morphology,image background processing,image edge expansion,image rotation and image noise are used to expand the numbers of samples from 384 to 12288,and then used the VGG16 model based on convolutional neural network to classify the samples.The accuracy rate can reach more than 94%.This kind of identification method has strong practical value and promotion prospects in oil and gas exploration field.