AI for DIGILAB: A New Concept in Digital Infrastructure for Heritage Materials Research

AI for DIGILAB：遗产材料研究数字基础设施的新概念

• 批准号: AH/T013184/1
• 负责人: Haida Liang
• 金额: $ 10.27万
• 依托单位: Nottingham Trent University
• 依托单位国家: 英国
• 项目类别: Research Grant
• 财政年份: 2020
• 资助国家: 英国
• 起止时间: 2020 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=AH%2FT013184%2F1
• 关键词: AI; DIGILAB; New; Concept; Digital

Over the last 20 years, digital imaging or digitisation of collections has become the norm within museums and archives. However, so far, it has mainly been focussed on recording what humans can see with their eyes, that is, colour RGB images and sometimes laser scanning of 3D objects. Growing interest by digital humanities scholars and medievalists in advanced imaging techniques and the new layers of information they can uncover affirms that the curatorial, art historical, and historical fields are receptive to a concept that has been explored by heritage scientists for several decades. We propose that the material composition of heritage objects analysed through various modalities of imaging spectroscopy such as reflectance spectral imaging and macro X-ray fluorescence (MA-XRF) scanning may be incorporated into digitisation campaigns to deliver a disruptive transformation in arts and humanities scholarship related to heritage. Since each material combination has its unique spectrum, imaging spectroscopy, depending on the modality, records to a greater or lesser extent, the material makeup (e.g. the pigments, dyes, binders, substrates) of an object. These added layers of information about heritage objects can lead to new insights and narratives about their creation, history of trade and cultural influences, and can impact significantly on conservation and preservation decisions. In addition, reflectance spectral imaging in the visible naturally gives the most accurate colour images thus removing the need for recording colour images. Nottingham Trent University's (NTU) ISAAC research group has made the first step in automatic collection of high spatial resolution reflectance spectral images of tens of square metres of wall paintings. Automatic data collection increases significantly the rate of data generation necessitating an automatic tool to process and reduce the data. While ML/AI has been used mostly in searching and organising digital content in the sector, the ISAAC team has pioneered their use in large scale heritage materials analysis. In addition to a new bespoke digital tool, we propose a new concept in digital research infrastructure through making the tool available to users remotely. The European Research Infrastructure for Heritage Science is divided into 4 platforms of operations: archives (ARCHLAB), mobile laboratory (MOLAB), fixed laboratory (FIXLAB) and digital laboratory (DIGILAB). While the first three platforms are well established, DIGILAB is in the concept phase, but offers opportunities for transformation in terms of access to digital tools and resources. Here we propose a model where DIGILAB functions as a data analysis lab where the user is helped remotely with their data analysis. The ML code will automatically process the image cubes into materials cluster maps and the experts will examine the results before releasing it to the users. A user interface and a visualisation add-on will also be developed to allow the user to view the outputs in a user-friendly manner. The remote access to digital lab aspect of the project aims to lower the barriers collections and scholars face in unlocking potentially relevant information encoded in the identity and distribution of materials used in the creation of heritage objects. Three varied case studies, each chosen for their different data-related challenges, will also serve to demonstrate and address issues in the workflow, starting from scientific data collection, then the new concept of DIGILAB for data reduction leading to material identification using complementary spectroscopic techniques, to finally address the research questions in history and conservation.

在过去的20年里，馆藏的数字成像或数字化已经成为博物馆和档案馆的常态。然而，到目前为止，它主要集中在记录人类眼睛所能看到的东西，即彩色RGB图像，有时也会对3D物体进行激光扫描。数字人文学者和中世纪学者对先进的成像技术和他们可以发现的新信息层的兴趣日益浓厚，这证实了策展、艺术史和历史领域正在接受遗产科学家已经探索了几十年的概念。我们建议，通过各种成像光谱（如反射光谱成像和宏观x射线荧光（MA-XRF）扫描）分析遗产物体的物质成分，可以将其纳入数字化活动，以实现与遗产相关的艺术和人文学术的颠覆性转变。由于每种材料组合都有其独特的光谱，因此成像光谱学根据其形态或多或少地记录了物体的材料组成（例如颜料、染料、粘合剂、基材）。这些关于遗产物品的附加信息层可以导致对其创作，贸易历史和文化影响的新见解和叙述，并可以对保护和保存决策产生重大影响。此外，反射光谱成像在可见自然给出最准确的彩色图像，从而消除了需要记录彩色图像。诺丁汉特伦特大学（NTU） ISAAC研究小组在自动收集数十平方米壁画的高空间分辨率反射光谱图像方面迈出了第一步。自动数据收集显著提高了数据生成的速度，因此需要一个自动工具来处理和减少数据。虽然ML/AI主要用于搜索和组织该领域的数字内容，但ISAAC团队率先将其用于大规模遗产材料分析。除了一个新的定制的数字工具，我们提出了一个新的概念，在数字研究基础设施，使工具可供用户远程使用。欧洲遗产科学研究基础设施分为四个操作平台：档案（ARCHLAB），移动实验室（MOLAB），固定实验室（FIXLAB）和数字实验室（DIGILAB）。虽然前三个平台已经建立良好，但DIGILAB仍处于概念阶段，但在获取数字工具和资源方面为转型提供了机会。在这里，我们提出了一个模型，其中DIGILAB作为一个数据分析实验室，用户可以远程帮助他们进行数据分析。机器学习代码将自动将图像立方体处理成材料聚类图，专家将在将结果发布给用户之前检查结果。还将开发用户界面和可视化附加组件，以便用户以用户友好的方式查看输出。该项目的远程访问数字实验室方面旨在降低收藏者和学者在解锁遗产物品创建中使用的材料的身份和分布中编码的潜在相关信息时所面临的障碍。三个不同的案例研究，每个都选择了不同的数据相关挑战，也将有助于展示和解决工作流程中的问题，从科学数据收集开始，然后是DIGILAB的新概念，用于数据减少，导致使用互补光谱技术进行材料鉴定，最后解决历史和保护方面的研究问题。

项目成果

Deep learning for the extraction of sketches from spectral images of historical paintings

• DOI: 10.1117/12.2593680
• 发表时间: 2021-06
• 期刊: Current Opinion in Oncology
• 影响因子: 3.4
• 作者: Qunxi Zhang;Shanshan Cui;Lu Liu;Jiaxin Wang;Jun Wang;Erlei Zhang;Jinye Peng;S. Kogou;Florence S. Liggins;Haida Liang

A new approach to the interpretation of XRF spectral imaging data using neural networks

使用神经网络解释 XRF 光谱成像数据的新方法

• DOI: 10.1002/xrs.3188
• 发表时间: 2020
• 期刊: X-Ray Spectrometry
• 影响因子: 1.2
• 作者: Kogou S

Maps and Colours - A Complex Relationship

地图和颜色 - 复杂的关系

• DOI: 10.1163/9789004467361_014
• 发表时间: 2024
• 作者: Liang H

Hyperspectral imaging solutions for the non-invasive detection and automated mapping of copper trihydroxychlorides in ancient bronze

用于古代青铜中三羟基氯化铜的非侵入性检测和自动绘图的高光谱成像解决方案

• DOI: 10.1186/s40494-022-00765-8
• 发表时间: 2022
• 期刊: Heritage Science
• 影响因子: 2.5
• 作者: Liggins F