基于双解码器网络的岩心CT图像分割

Core CT image segmentation based on dual decoder network

在石油地质领域,分析岩心内部微观孔隙结构的形态分布、特征参数,对研究油气资源的渗流特性和储集性能具有重要意义。岩心CT图像具有噪点多、比度低、亮度不均匀的特点,目前实际工程运用的孔隙提取方法,仍然存在着需要大量人工交互且分割精度较低的问题。针对这些问题,本文提出了基于双解码器网络的分割方法,构建了岩心CT图像分割数据集,用图像预处理网络分支,辅助训练图像分割网络。实验结果表明,本方法的模型参数量仅有33.3 MB,像素精度PA能达到91.41%,平均检测交并比MIoU能达到85.32%,具有模型小、推理速度快、分割精度高的优点。

In the field of petroleum geology,analyzing the morphological distribution and characteristic parameters of the microscopic pore structure inside the core is of great significance to study the seepage characteristics and reservoir performance of oil and gas resources.Core CT images have the characteristics of high noise,low ratio and uneven brightness.The pore extraction methods used in practical engineering still have the problems of requiring a lot of manual interaction and low segmentation accuracy.In order to solve these problems,this paper proposes a segmentation method based on dual decoder network which uses image preprocessing network branches to assist in training image segmentation network.The number of model parameters in this method is only 33.3 MB.Besides,we construct a core CT image segmentation dataset.Experimental results show that the pixel accuracy PA can reach 91.41%,and the average detection intersection and merge ratio MIoU can reach 85.32%,which prove that our method has the advantages of small model,fast inference speed,and high segmentation accuracy.