基于深度学习建立模型去除工业CT伪影的技术研究

Research on The Technology of Removing Artifacts in Industrial CT Based on Deep Learning Modeling

针对高压涡轮叶片工业CT检测过程中伪影干扰严重的问题,采用一种基于深度学习技术的伪影去除算法,建立去伪影自编码器模型,实现工业CT截面图像伪影的修正。通过试验证明,采用本算法训练获得的修正模型,能够有效去除高压涡轮叶片CT检测截面图像上的伪影干扰,并具有可推广适用性。

This article aiming at the serious influence of artifacts in the process of industrial CT detection of highpressure turbine blades,an artifact removal algorithm based on deep learning technology is used to establish a de-artifact autoencoder model to correct the artifacts of industrial CT cross-section images.It is proved by experiments that the correction model obtained by the training of this algorithm can effectively remove the artifact interference on the CT inspection cross-sectional image of the high-pressure turbine blade,and has generalized applicability.