This paper outlines a new framework for applying Structure-from-Motion (SfM) to challenging, feature-poor environments such as those observed during AGR fuel channel inspection. Deriving structural information from AGR inspection footage is challenging due to several key issues; lack of discriminative salient features within the channel, inconsistency in lighting during the inspection process, lack of textural information within the channel and noise from the inspection equipment. This presents difficulties to techniques such as SfM due to its reliance on finding and reliably tracking a set of robust features from multiple viewpoints. This paper introduces the first use of an incremental 3-D reconstruction framework which can produce reconstructions of footage obtained within a nuclear reactor. It approaches this issue by introducing a novel correspondence searching methodology which can operate within feature-poor environments by utilising a constrained, iterative threshold matching technique to obtain robust feature matches. This paper demonstrates the approach using two datasets; laboratory footage obtained from an experimental setup emulating a small sub-section of the channel and in-core inspection footage of AGR fuel channels.
|Number of pages||10|
|Journal||Nuclear Engineering and Design|
|Publication status||Accepted/In press - 2 Dec 2019|
- AGR cores
- 3-D visualization
- advanced gas-cooled reactor (AGR)