Prof Dr Benedikt Hensel

"Hazor Digital" and its research objectives

Digital methods have been tried and tested in archaeology for years and are already established in some areas - including geoinformation systems, databases, digital photography, digital 3D documentation and 3D scanning, virtual reconstructions, retro-digitisation of old analogue documentation (diaries, plans, drawings) as well as statistics, machine learning, artificial intelligence and agent-based simulations. At the same time, these methods in archaeological projects are often organised as individual case solutions: they address individual aspects (planum, profile, object, find inventory, image data, evaluation), are tailored to specific methods and only interact to a limited extent as an integrated overall process of field research. This is precisely where the Oldenburg research project "Hazor Digital" comes in and addresses two central aspects of digital field archaeology:

1. a data information system - a (field) archaeological processing pipeline

Theaim is to develop an integrated (field) archaeological processing pipeline that maps the excavation project as a continuous, multimodal digitised process - from excavation planning and organisation to ongoing field documentation, AI-supported evaluation, geovisualisation and presentation of the results as a classic publication in a web-based research environment, as well as a solution for long-term archiving of the research data. Data streams (text, photo, 3D, surveying, GIS) are consistently merged: imaging methods, georeferenced data input and mobile recording during the excavation (tablets/smartphones) feed directly into a common data information system, which will be developed for the requirements of a large-scale, stratigraphically complex excavation as part of "Hazor Digital". At the same time, the FAIR principles (Findable, Accessible, Interoperable, Reusable) are consistently applied: Data, metadata and workflows should be standardised, documented and versioned in such a way that they remain findable, accessible, interoperable and reusable in the long term - including clear responsibilities, interfaces and quality assurance routines in the field and in post-processing.

For the Oldenburg co-led excavation in Hazor (Israel), this means the transition from handwritten documentation and manually created plans (stone-based measurements) to a digital excavation documentation system in which daily documentation and excavation photography are recorded directly in the field - supplemented by regular, i.e. daily, three-dimensional and georeferenced documentation of the actively processed sections.

2. AI-supported, (partially) automated evaluation of the field archaeological features and finds in the data information system and pipeline

A central core component of the processing pipeline is the (partial) automation of the transfer of documentation products into analysable plan and analysis data: Marked features are assigned from orthomosaics on a stratum/epoch basis and converted into georeferenced vector data; image segmentation, vectorisation, georeferencing and (partially) automated feature annotation are integrated as key processes in a GIS pipeline and interlinked with the excavation database. In addition - as a particularly important research focus - AI-supported automatic image segmentation and object recognition in archaeological contexts, which has so far only reached a limited level of maturity and therefore requires systematic training, validation and iterative improvement. In particular, the focus is on automated wall recognition, stone detection and structure detection (e.g. wall crowns, stone settings, building remains, edge courses), which are typically generated by draughtsmen and manual surveying in classical excavation practice. The aim is to gradually transfer this previously highly manual translation work - from the findings in the planum to a representation that can be planned and analysed - into a reliable, quality-assured digital workflow that remains documented, reproducible and stratigraphically controllable.

At the same time, imaging optimisations (e.g. HDR) for elusive feature phenomena (soil discolouration, clay brick remains, profile layers) are being tested and coupled with 3D models and orthomosaics in order to support the evaluation analytically. On this basis, 3D models are to be stored in an accessible infrastructure so that they can be used as a visual and analytical reference during the fieldwork and evaluation phases (including survey data and feature/time markers). In the future, digital twins of the excavation - differentiated according to strata - can be created from this, which provide a basis for reconstructions and 3D visualisations at different historical points in time.

In addition, a key objective is to publish results more quickly and in a connectable way: via an interactive web hub with clear standards of data organisation, documentation and reusability in order to reach the specialist world quickly and at the same time provide a consistent multimodal basis for AI-supported analyses.

Fig.: Exemplary process plan of an archaeological excavation project