Advanced video analysis for automated feature identification on Special Nuclear Materials (SNM) packages

用于特殊核材料 (SNM) 封装自动特征识别的高级视频分析

• 批准号: 2897614
• 金额: --
• 依托单位: University of Strathclyde
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2023
• 资助国家: 英国
• 起止时间: 2023 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2897614
• 关键词: Advanced; video; analysis; automated; feature

Sellafield Ltd is responsible for the storage of Special Nuclear Materials (SNM) that are a legacy of 60 years of reprocessing activities on the Sellafield site. To provide confidence that SNM packages remain safe for continued storage within Sellafield's stores, it is necessary to closely monitor the exterior surface of the packages whilst they are in their storage location (in-situ). Such monitoring will inform the selection of packages for ex-situ inspection.The aim of this research is to develop new image and video processing algorithms which can be used to analyse existing and future SNM inspection videos to automatically detect and quantify the condition of each SNM package that undergoes inspection. Specifically, techniques will be designed to detect and quantify the following characteristics of SNM packages:-Dimensions of any dents, scratches or scuffs-Evidence of cracking-Original manufacturing defects-Evidence of corrosion/colour changeIf successful, new methods will then be designed to perform batch analysis of all SNM containers to undergo inspection with the aim of identifying those at most and least risk of requiring ex-situ inspection and, possibly, some form of re-processing and/or re-packaging. The explicability of the feature detection algorithms will also be assessed with the aim of designing explicable AI based solutions.

塞拉菲尔德有限公司负责储存特殊核材料（SNM），这些材料是塞拉菲尔德场地60年后处理活动的遗留物。为了确保SNM包装在塞拉菲尔德的商店内继续储存时保持安全，有必要在包装在其储存位置（原位）时密切监测包装的外表面。这种监测将通知选择的包为ex-situ inspection.本研究的目的是开发新的图像和视频处理算法，可用于分析现有的和未来的SNM检查视频自动检测和量化的条件下，每个SNM包进行检查。具体而言，将设计技术来检测和量化SNM包装的以下特征：-任何凹痕、划痕或磨损的尺寸-开裂的证据-原始制造缺陷-腐蚀/颜色变化的证据如果成功，则将设计新的方法来对所有SNM容器进行批分析，以进行检查，目的是确定需要异地检查的风险最大和最小的容器，以及可能需要某种形式的再加工和/或再包装的容器。还将评估特征检测算法的可解释性，以设计可解释的基于AI的解决方案。