An Enhancement Algorithm for Concrete Imaging Based on WDCT Network
-
摘要: 原位加载CT扫描可以研究混凝土在不同实验条件下的损伤演化过程，对于混凝土材料研究具有重要意义。在混凝土原位加载的过程中，原位加载设备的遮挡和采样时间的限制导致了有限角度和稀疏角度CT重建问题，传统算法在数据不完备的情况下会导致严重的图像伪影。本文提出WDCT网络增强ART-TV重建图像方法，构建混凝土模拟数据集验证该方法并与经典U-Net增强FBP重建图像方法进行比较。研究结果表明本文的方法在有效降低图像伪影的同时，能够对混凝土内部的石料和裂缝比较准确地成像，图像的客观指标得到大幅提升。Abstract: Field loading tests with CT can research the damage evolution process of concrete under different experimental conditions, which is of great significance for the research of concrete materials. In the process of field loading tests, the occlusion of field loading equipment and the limitation of sampling time lead to the limited-angle problem and the few-view problem. Traditional algorithms can cause serious image artifacts with incomplete data. We propose to use WDCT network to enhance ART-TV reconstruction images and build a concrete simulation data set to compare the difference between our method and the classic method enhancing FBP reconstructed images with U-Net on the subjective effect and objective indicators. The experiment results show that our method can accurately reconstruct stones and cracks inside the concrete and effectively reduce image artifacts. The objective indicators of images are greatly improved.
-
Keywords:
- industrial CT /
- concrete inspection /
- limited-angle /
- few-view /
- deep learning
-
计量
- 文章访问数: 585
- HTML全文浏览量: 22
- PDF下载量: 54
下载:
