首页|期刊导航|计算机与数字工程|基于改进DeepLabv3+的野外露头区岩石裂缝分割算法

基于改进DeepLabv3+的野外露头区岩石裂缝分割算法OA

An Improved DeepLabv3+based Rock Fracture Segmentation Algorithm for Outcrop Areas

中文摘要英文摘要

探究野外露头区岩石裂缝的分布对于裂缝性油气储藏的地质勘测具有重要的指导作用.针对传统图像处理方法难以准确提取裂缝的问题,论文提出了一种改进的DeepLabv3+野外露头区岩石裂缝分割算法,该算法在编码器部分使用更轻量化的Mobilenetv2网络进行提取特征减小了模型参数;同时使用密集连接的方式优化ASPP模块,在增大感受野的同时增强了分割裂缝边缘的能力;使用Focal loss损失函数修正数据中裂缝和不同岩石背景不平衡的问题.在野外露头区岩石裂缝数据集进行实验结果表明,改进的DeepLabv3+算法比原算法具有更高的裂缝分割精度,Mean accuracy提高了5.3%,MIoU提高了1.6%,对于裂缝的分割更准确.

Exploring the distribution of rock fractures in outcrop areas in the field has an important guiding role for geological exploration of fractured oil and gas reservoirs.Aiming at the problem that traditional image processing methods are difficult to accu-rately extract fractures,this paper proposes an improved DeepLabv3+outcrop rock fracture segmentation algorithm.This algorithm uses a lighter weight Mobilenetv2 network in the encoder section to extract features to reduce model parameters.At the same time,the dense connection method is used to optimize the ASPP module,which enhances the ability to segment the edge of the fissure while increasing the receptive field.The Focal loss function is used to correct the imbalance between fractures and different rock backgrounds in the data.Experimental results on rock fracture data sets in outcrop areas show that the improved DeepLabv3+algo-rithm has higher fracture segmentation accuracy than the original algorithm,with an increase in Mean accuracy of 5.3%and an in-crease in MIoU of 1.6%,making it more accurate for fracture segmentation.

王婷婷;张梦柳;赵万春

东北石油大学电气信息工程学院 大庆 163318东北石油大学电气信息工程学院 大庆 163318东北石油大学石油工程学院 大庆 163318

信息技术与安全科学

图像语义分割岩石裂缝识别深度学习DeepLabv3+

image semantic segmentationidentification of rock fracturesdeep learningDeepLabv3+

《计算机与数字工程》 2026 (3)

634-639,6

10.3969/j.issn.1672-9722.2026.03.009

评论