摘要
由于CT序列图受到成像物体和设备精度影响,传统的CT序列图分割方法需要大量的图像运算,提取的裂缝容易受到噪声的影响。对此,将CT序列图看作一个连续变化的视频序列,使用光流神经网络模型进行帧间变化检测,通过区域聚类和光流阈值实现目标的分割。实验结果表明,该方法很好地解决了岩心CT序列图的裂缝分割,并提高了分割的自动化水平。
Because CT image sequence is affected by the imaging objects and the accuracy of equipment,traditional segmentation methods of CT image sequence require a lot of image operations,and the extracted cracks are vulnerable to noise.In this paper,CT image sequence is regarded as a continuously changing video sequence,and the optical flow neural network model was used to detect the change of the inter-frame.The new target was updated by optical flow threshold and region clustering.The experimental results show that this method solves the crack segmentation of core CT sequence diagram and improves the automation level of segmentation.
作者
谢杨灏
滕奇志
Xie Yanghao;Teng Qizhi(Institute of Image Information,College of Electronics and Information Engineering,Sichuan University,Chengdu 610065,Sichuan,China)
出处
《计算机应用与软件》
北大核心
2021年第1期211-216,共6页
Computer Applications and Software
基金
国家自然科学基金项目(61372174)。
关键词
数字岩心
CT序列图
光流网络
裂缝分割
深度学习
Digital core
CT image sequence
Optical flow network
Crack segmentation
Deep learning
