1Every year maritime archaeologists around the globe are faced with the daunting task of having to record, interpret and publish hundreds if not thousands of individual ship timbers. After in situ documentation entire wooden wrecks are frequently disassembled, recovered and subsequently recorded frame by frame, plank by plank, before either being conserved, stored or discarded. This in-depth study of individual timbers is often crucial to understanding a ship’s construction, and the resulting timber catalogue serves as a final, detailed visual record of the vessel’s numerous individual elements.
2Current methods for timber recording, be they digital or analogue, 3D or 2D, are generally quite accurate and detailed, but unfortunately also very time consuming. In practice the recording of a single wreck may take many months and in some cases years to complete. On time-critical projects with large quantities of timbers, such as many development-led excavations, these methods are therefore often simply untenable.
3In this paper we present a novel approach to timber recording which addresses this issue. We begin by briefly outlining our use case and recording objectives. Next, we evaluate whether the current best practice method for timber recording, namely contact digitising, fulfils these objectives. We then present our proposed methodology and conclude by comparing the outcomes of both methods.
4Different facets of our proposed methodology were gradually refined between 2015 and 2018 over the course of seven development-led shipwreck documentation projects. The studied shipwrecks include the 15th-century Ijssel cog, punt and barge wrecks excavated by archaeology company ADC ArcheoProjecten in the River IJssel in the Netherlands (Van Damme 2018) and two 18th-century shipwrecks excavated by the Landesamt für Kultur-und Denkmalpflege (LAKD) Mecklenburg-Vorpommern in the Bay of Greifswald in northern Germany (Auer et al. 2018 and Auer in this volume).
5In this paper, we will focus on the most recent implementation of the methodology, as used on the Wismar Big Ship project (Ditta, Auer in this volume). The 12th-century Wismar Big Ship measures 19.8 by 5.3 m and was excavated by the LAKD in the harbour of Wismar in the winter of 2017. After excavation, a total of 240 diagnostic timber elements were selected to be recorded using our methodology: 106 two-sided timbers (or parts thereof) and 134 four-sided elements.
6Despite the development-led nature of these projects, the intention was to adhere to quite demanding standards in terms of timber recording. Specifically, the recording objectives were: a) to be accurate, detailed and three-dimensional; b) to allow for 3D printing of the ship timbers and 3D modelling of the wreck; c) to allow for the creation of a 2D timber catalogue for final publication; and d) to be time- and cost-efficient.
7Whereas traditionally ship timbers have been recorded in 2D using scaled drawings or full-sized tracings, over the past 15 years 3D contact digitising has gradually emerged as the best practice method for ship timber recording. Contact digitising involves tracing all significant features and outlines of a timber using the pen of a coordinate measuring device, which tracks and records the precise 3D location of the penpoint, generally to sub-millimetre accuracy. Typically the traced 3D coordinates are recorded in real time as 3D polyline wireframes in the 3D CAD software Rhino, with different layers in Rhino representing different feature types. The best-known and most widely used tool for this approach is the FaroArm, though other coordinate measuring devices such as the wireless Creaform HandyPROBE have also been used (Jones 2015).
8Within the field of maritime archaeology the FaroArm was first used at Mystic Seaport in the US in the mid-1990s to record ship’s models (Starr 1996). For timber recording, use of the FaroArm and Rhino 3D was pioneered by the National Museum of Denmark in 2001 to record 56 timbers of the Kolding cog (Hocker 2003; Jones 2015, p. 143-147). Since then the approach has seen widespread adoption on numerous shipwreck projects, mainly in Europe and Turkey, including the Roskilde wrecks, the Newport ship, the Drogheda boat, the Doel cog project, several of the Yenikapı shipwrecks and Arles-Rhône 3 (Jones 2015, p. 264-282; Kocabas 2015; Lenaerts et al. 2011; Marlier et al. 2017; Schweitzer 2012).
9Compared to traditional 2D recording methods, contact digitising presents a clear step forwards in terms of accuracy, while simultaneously producing a full three-dimensional record of each timber. By using similar layering conventions in Rhino, the digital 3D timber files can easily be understood at a glance by researchers on different projects. Three-dimensional layers can be switched on or off to create 2D views of each face of a timber for publication, and more importantly, the 3D polyline wireframes can be modelled into 3D solids. This allows researchers to reassemble the wrecks digitally in CAD software, or to 3D print the timbers and reassemble them physically. As such, contact digitising meets our three first recording objectives: a) it is accurate, detailed and three-dimensional; b) it allows for 3D printing and 3D modelling; and c) it allows for the creation of a 2D timber catalogue.
10The final question then is: d) is contact digitising also time- and cost-efficient? A first factor impacting cost-efficiency is the cost of the recording hardware. While prices vary depending on the model chosen, on average a new FaroArm retails at around €30,000 (Toby Jones, personal communication). The second and most significant factor influencing cost-efficiency is time-expenditure: the more time is needed to record the timbers, the more money will be spent on staff wages and other project operating costs, such as location rental and staff accommodation. In order to get an estimate for time-expenditure, we base ourselves on figures reported by Toby Jones for contact digitising on the Newport ship project, which is considered to be one of the most efficient implementations of this methodology. Jones has estimated that, on average, a FaroArm user with at least one month of experience would need about four hours to record a plank and six hours to record a frame. In addition, once the timbers had been recorded, converting the 3D polyline wireframes into 3D solids for CAD modelling or 3D printing required, on average, a further two hours per timber (Toby Jones, personal communication).
11Based on these estimates, assuming we were able to hire an experienced FaroArm user (and supporting staff to clean, photograph and make written descriptions of the timbers) working ten hours per day, contact digitising of 106 two-sided timbers and 134 four-sided timbers on the Wismar Big Ship project would have taken roughly 123 days to complete, or an average of about two timbers recorded per day. Since in this approach the main bottleneck is the hardware itself, the only way to speed up recording would be to purchase additional FaroArms, which, as we have seen above, are rather costly. Finally, after recording, an additional 480 labour hours would be needed to convert the 3D polylines into digital solids for 3D printing and modelling.
12Now consider the following: on the Wismar Big Ship project, the actual underwater excavation and underwater recording of the wreck, as well as mobilisation, demobilisation and timber transport, took a team of six archaeologists working eight hours per day a mere 32 days to complete. As such, it seems that for most development-led projects the added scientific value of recording all timbers using contact digitising is simply not proportional to the added time-expenditure and project costs associated with this method. This likely explains why in maritime archaeology contact digitising, despite being considered the best practice method for timber recording, has so far remained inaccessible to all but the best-funded research projects, or projects relying on free volunteer or intern labour, with access to free or low-cost lent-out contact digitisers. Unfortunately, the method is simply too slow and therefore too costly for most short-term development-led projects.
13The timber recording methodology we propose consists of several phases. After cleaning, timbers are first scanned in 3D. In practice, a wide range of 3D scanning methods can be used to make the timber scans: in the case of the Wismar Big Ship project, we opted to use an Artec Eva 3D scanner. After processing, the resulting scans are then imported into the CAD software Rhino 3D, where the timbers’ various important features are annotated directly onto the timber’s digital copy. Finally, scaled 2D technical line drawings and textured renders of all faces of each timber are then exported from Rhino using an automated workflow, in order to produce the wreck’s timber catalogue.
14The Artec Eva is a handheld structured-light 3D scanner which records both object geometry and texture (colour). Prior to scanning, several factors must be taken into consideration. In order to avoid interference with the structured-light system, scanning should be carried out indoors, out of direct sunlight. Reflections can likewise lead to errors, so after cleaning, timber surfaces should be blotted dry with a towel. Finally, the scanner tracks its location using both geometry and texture information, so it is advised to place the timbers on a surface that contains both distinctive geometry and texture features: in our case, simple wooden pallets proved to be an ideal background for scanning.
15In general, the workflow then consists of making several overlapping scans of each timber, and subsequently combining these scans into a single model using the Artec Studio software. During recording, the Eva was connected to a laptop, thereby allowing the researchers to move freely around the timber, and to see the 3D model being captured in real time on the computer screen (fig. 1). Sturdy four-sided timbers are the easiest to record: they can be placed directly on the pallets and recorded in two scans, one per side. Flexible planks with thin edges are more problematic: they may change shape when turned over, and their edges provide little overlap between the inboard and outboard scans. In order to overcome this, pins can be inserted along the edges of the plank to provide stable reference points between two scans, or the planks can be suspended above the recording surface (through staves inserted in treenail holes), allowing the plank to be recorded in a single scan (fig. 2).
Fig. 1: Timber scanning setup using the Artec Eva structured-light 3D scanner
(photograph T. Van Damme)
Fig. 2: Planks
Generally, planks will bend slightly when turned over, meaning that the 3D geometry of the plank changes in between two scans. This may lead to reconstruction errors when merging the outboard and inboard scans of a plank. In order to avoid this issue, planks can be suspended above the recording surface, thereby allowing the user to record the entire plank in a single scan.
(photograph J. Auer)
16While one person was scanning, another person was responsible for timber processing on a custom-built PC: this parallel workflow allowed us to avoid waiting times between tasks. In terms of processing, in Artec Studio the background of each scan is first removed. The individual scans are then aligned, either automatically or manually using reference points. Following successful alignment, the “Global Registration” command is used to optimize the alignment of the individual scans and to combine them into a single model of the timber. After removing noise and unwanted objects from the scan, the “Sharp Fusion” command is used to create a watertight mesh, which is then simplified and textured.
17The resulting textured mesh is an objective digital 3D reproduction of the physical timber, but as archaeologists we need to go one step further and actually interpret the timber. To do this, we first save the textured mesh in OBJ format and import it into Rhino 3D. In Rhino, we use a layering convention similar to that used for contact digitising, with different layers representing different feature types such as wood grain, treenails, nails, repairs, tool marks and intentional markings. Whereas with contact digitising these features are physically traced using the pen of a coordinate measuring device, in our approach the features are traced digitally using Rhino’s “PolylineOnMesh” command, which allows users to draw 3D polylines directly onto the textured mesh (fig. 3). During this annotation process we keep the physical timber next to us, and researchers go back and forth between examining the physical timber and then tracing all significant features onto the timber’s digital copy. A notable difference with contact digitising is that we do not need to trace all the timber’s edges, since these are already stored in the mesh geometry. In addition to tracing all significant features, we also produce cross sections of the timbers at intervals using the “Section” command, and we draw polylines connecting treenail ends in order to show the direction in which treenails are inserted. Finally, during this annotation phase researchers also make a written description of the timber and take detail pictures of important features.
Fig. 3: Rhino 3D screenshot
Rhino 3D screenshot, showing timber features traced on the textured mesh as 3D polylines (left) and Rhino layers representing different feature types (right).
18In order to produce a 2D timber catalogue entry for each timber, we take advantage of Rhino’s various “Drafting/Layout” and “Rendering” tools. Using the “Drafting/Layout” tool we create standard templates for the timber catalogue, in which each face of the timber (two faces for planks, four faces for frames) is automatically shown at scale 1:10 on a (digital) sheet of paper ranging in size from A4 to A0 (depending on the size of the timber). After creating several standard layout templates, the user simply has to place the timber in the right position in 3D space in Rhino and choose the most appropriate template for any given timber. The scaled timber drawings are then exported in PDF format both as “Technical” and “Rendered” views. The “Technical” render automatically converts the mesh into a line drawing in which timber outlines and geometric edges are shown in black, and the traced features are shown in different colours corresponding to the layers on which the features were traced. The “Rendered” view shows a full colour render of the textured mesh (fig. 4).
Fig. 4: Example of a four-sided timber catalogue entry, showing both the “Technical” view (left) and “Rendered” view (right)
19The level of accuracy and detail of the Annotated Scans Method will depend largely on the 3D scanning method used to produce the original timber scans. The Artec Eva used in our case study has a stated accuracy of up to 0.1 mm, and a resolution of up to 0.5 mm (Artec 2019). Given that traditional shipbuilders likely were not working at sub-millimetre levels of precision, this should be more than enough for most maritime archaeological applications. Since the result of the initial scanning phase is a textured mesh of each timber, these 3D solids could immediately be used for 3D modelling and printing. On the Wismar Big Ship project, a preliminary CAD model assembly of the wreck’s framing timbers was produced on the fly as new scans were processed, and after scanning was completed, all timbers were 3D printed at scale 1:20 and reassembled using metal wire. The resulting model was again scanned in 3D and a tentative CAD reconstruction of the original vessel was made (Ditta, Auer in this volume). By annotating the meshes in 3D in Rhino we were able not only to record but also interpret the ship timbers, and Rhino’s “Drafting/Layout” tools proved invaluable in order to automatically produce 2D timber catalogue entries of all faces of each timber. As such, the annotated scans method meets our three first recording objectives: a) it is accurate, detailed and three-dimensional; b) it allows for 3D printing and 3D modelling; and c) it allows for the creation of a 2D timber catalogue.
20Once again, the final question then is: d) is the method also time- and cost-efficient? In terms of hardware, the most significant cost is the 3D scanner. At the time of writing, a new Artec Eva costs either €12,330 or €13,700, depending on whether the user qualifies for an educational discount (Artec 2019). However, as discussed above, by far the most important factor determining the cost of a project is time-expenditure. Using the annotated scans method, on the Wismar Big Ship project a team of four people equipped with a single Artec Eva scanner, working ten hours a day, were able to clean, scan, annotate, photograph and describe 240 timber elements in a mere 37 days. This adds up to an average of 6.5 timbers per day. Interestingly, one or two people scanning and processing timber scans were able to complete eight to ten timber scans a day, meaning that in our approach the main bottleneck was not the scanning hardware itself, but rather the speed at which the other researchers could subsequently annotate and describe the timbers. Since the final output of the scanning phase is a textured mesh, no additional manual work was required to convert our scans into digital solids for 3D modelling and printing. Finally, after recording was completed, it took a single person ten days to produce a detailed timber catalogue of all 240 timber elements.
21It should be noted that the Wismar Big Ship project had several advantages. Firstly, three out of four team members had prior experience with the Artec Eva scanner and the annotated scans method while working on the shipwrecks excavated in the Bay of Greifswald the year before. Secondly, the team was highly qualified in terms of general 3D scanning experience and know-how, meaning that workflows could quickly be adapted to overcome the challenges that inevitably come up while scanning timbers. A less experienced team would likely have hit more obstacles over the course of the project, resulting in significantly longer scanning times. Nevertheless, the project also suffered several setbacks: our Artec Eva scanner was faulty for two weeks, resulting in significantly longer processing times; two team members were sick for a week; and at one point the workplace flooded, which resulted in a day lost to cleaning. As such our case study does not represent the “best case scenario”, and we are confident that we can further improve on time-efficiency in future projects.
22The annotated scans method matches, and in several regards surpasses, the current best practice method for timber recording. Firstly, whereas contact digitising only stores the information considered important by the researcher at the time of recording, our method results in an objective digital copy of every geometric and colour detail visible on the original timber (the textured mesh), which is then supplemented by the archaeologists’ interpretation of the timber during the annotation phase. Secondly, in contrast to the simplified 3D solids produced with contact digitising, the 3D scans made using our method contain significantly more geometric detail, resulting in more accurate models for 3D printing or CAD modelling. Most importantly, however, the annotated scans method is considerably more cost-efficient than contact digitising, thanks in part to reliance on cheaper hardware, but mostly due to substantial reductions in time-expenditure.