Photogrammetric Three-dimensional Modeling and Printing of Cetacean Skeleton using an Omura's Whale Stranded in Hong Kong Waters as an Example

The bone morphology of a semi-degreased baleen whale was documented by photogrammetry with a DSLR camera to generate three-dimensional (3D) models by computer, which were 3D printed as half-sized replicas of the original for display and educational purposes.

The preparation of cetacean, in particular baleen whale, skeletons presents a great challenge due to their high lipid content and uncommon size. Documentation of the skeletal morphology is important to produce accurate and reliable models for both research and educational purposes. In this paper, we used a 10.8-meter long Omura’s whale stranded in Hong Kong waters in 2014 as an example for the illustration. This rare and enormous specimen was defleshed, macerated, and sun-dried to yield the skeleton for research and public display. Morphology of each bone was then documented by photogrammetry. The complex contour of the skeleton made automated photoshoot inadequate and 3 manual methods were used on bones of different sizes and shapes. The captured photos were processed to generate three-dimensional (3D) models of 166 individual bones. The skeleton was printed half-size with polylactic acid for display purposes, which was easier to maintain than the actual cetacean bones with high residual fat content. The printed bones reflected most anatomical features of the specimen, including the bowing out rostral region and the caudal condylar facet that articulated with Ce1, yet the foramina on the parieto-squamosal suture, which are diagnostic character of Balaenoptera omurai, and an indented groove on the frontal bone at the posterior end of the lateral edge were not clearly presented. Extra photoshoots or 3D surface scanning should be performed on areas with meticulous details to improve precision of the models. The electronic files of the 3D skeleton were published online to reach a global audience and facilitate scientific collaboration among researchers worldwide.

Cetacean strandings offer valuable opportunities to learn about their life history, biological health and profile, as well as the effect of anthropogenic actions to the ecosystem. Three-dimensional (3D) representation and modeling allow accurate representation of morphometric measurements that can be used for biomechanical calculations and give insights on various physiological behaviors¹. Morphological adaptations have allowed these animals to survive in the ocean, while some pathologies observed in stranded cetaceans could reveal their biological health and profile, anthropogenic and non-anthropogenic circumstances or cause of death^2,3. Bone lesion followed by traumatic collision may remain unhealed since the animals are required to swim continuously under tremendous underwater pressure⁴. In marine mammals, compression and non-fatal gas embolism may decrease blood supply to the bones and cause barotraumas⁵. Adverse bone remodeling may result in both pain and decreased spinal mobility that compromises their survival upon predation or other threats. Increases in the reporting of mortality and morbidity in cetaceans worldwide also indicated the possible decline in ocean health^6,7. Recognizing the importance of the ocean and the inextricable links between human health and cetacean and ecosystem health has led to the ‘One Ocean−One Health’ research paradigm⁸.

On March 31st 2014, an Omura’s whale (Balaenoptera omurai) stranded near Hung Shek Mun, Plover Cove Country Park, Hong Kong. It was a 10.8-meter adult female, and only a few of these species have been found in the Indo-Pacific region since it was first discovered in 2003⁹. The beaching of a whale of this size is not common in Hong Kong, therefore, this event presented an opportunity to preserve the skeleton for research and educational purposes. The beached animal was dissected and defleshed at the site of discovery, with the majority of external muscles and internal organs removed. Gross necropsy revealed that the carcass was in an advanced state of autolysis but had multiple deep lacerations crisscrossing the body, the most severe of them centered on the right pectoral fin with a deep transverse laceration extending throughout the bone, demonstrating a certain degree of confident linkage of entanglement evidence with an observed condition of the mortality. The skeletal remains were transported to a location in Lantau island by the Agriculture, Fisheries, and Conservation Department of the Government of the Hong Kong Special Administrative Region, where maggots were utilized to consume the soft tissues. The bones were degreased by water maceration for 2 months with manual scrubbing. Despite a significant amount of bone fat being extracted, the skeleton, in particular the skull and rib edges, remained brown in color. The residual fat was difficult to remove and if left untreated, would attract rodents, deteriorate, and render the specimen unfit for presentation. Even with perfect conservation conditions, animal bones can still be decomposed by various dust-inhabiting microorganisms¹⁰. It was decided to have the morphology of the semi-degreased skeleton digitally documented and then 3D printed with durable materials as a sanitary replica of the original.

Biological specimen 3D models can be generated by several means, including medical imaging, surface scanning, and photogrammetry. Medical imaging modalities like Computed Tomography (CT) and Magnetic Resonance Imaging produce multiplanar images that include both external and internal features, but lack color and texture. CT has been utilized to document the anatomy or pathology of flukes, baleen, and cranium of various species, revealing their unique adaptation in locomotion, foraging, and neurodevelopment^{11,12,13,14,15,16}. Surface scanning projects either laser or structure light on the object, where the pattern of reflection is converted to geometric data by trigonometric triangulation to generate a surface model. Photogrammetry records a series of slightly overlapping photos of the target. Either the camera rotates around the object, or the object is rotated on a turntable upon shooting. The process is repeated with different camera angles and heights before the object is turned over to capture the underside likewise. Photos are imported into a modeling software, which calculates location and distance of each feature in 3D space to produce point clouds. The geometric information is processed by triangulation of the point clouds to generate polygonal meshes, which can be edited and manufactured. 3D reconstruction can reflect accurate measurements of scanned surfaces and volumes¹⁷.

Photogrammetry was considered a suitable approach for 3D documentation of the Omura’s whale skeleton in view of its low equipment cost, adequate output quality, and flexibility in dealing with bones of largely variable sizes and shapes. For instance, the whale skull measured 2.6 meters, which made miniature methods like laser 3D surface scanning unfeasible. The equipment required for photogrammetry is readily accessible – only a digital camera with high capture resolution (>5 megapixels) and a modeling software, which is much cheaper than the optic or laser scanners for 3D surface scanning. In addition, 3D surface scanning requires the scanner to be connected to a reasonably high-performance computer during data collection, both of which require an independent power supply. 3D surface scanning is inapplicable when a power source is absent, for instance in case of very large specimens with limited transportability, or when the original whale carcass is to be scanned on-site. For photogrammetry, only a digital camera, a tripod, and a supporting apparatus such as a turntable are needed. Photogrammetry is therefore a more affordable option with high portability for small research groups to start with.

Digital models are transformed into physical products by 3D printing. Layers of melted poly-lactic acid (PLA) are stacked and solidified to reproduce the whale skeleton. The realistic replica, printed half-size, can be used for public display and educational purposes. For students and layman in general, touching anatomical models can help them appreciate the animal not only visually but also by sensation. For professionals like young clinicians and scientists, it can be difficult to understand complicated structures from 2D pictures¹⁸. Traditionally, biological specimens undergo plastination to become educational adjuncts, yet the process is quite complicated, resource-demanding, and time-consuming. Carcasses may possess biological hazards, and only one model is generated from each specimen. 3D documentation and printing offer interactive experiences that are more enjoyable than textbooks or virtual animations. Even virtual dissection cannot offer the benefits of tangible manipulation and is thus unpopular among students¹⁹. With 3D printing technology, multiple copies of a rare specimen can be replicated, held in hand, and studied closely from different angles, without undesirable odor or the fear of breaking them²⁰. The product can be customized, for example scaled down for easy manipulation or printed in different colors for aesthetic illustration. The 3D models can also be digitally edited to restore broken or missing parts which allows greater versatility. 3D documentation and printing also facilitate knowledge sharing among researchers. A skeletal remain can be digitally recorded, shared online, and printed on demand. Specimens can be “prototyped” and distributed overseas as a standard parcel instead of a biological sample, which requires special quarantine or legal documentation. Electronic 3D models comprising key metrics of the whale bones are also shared online with other institutes to facilitate scientific collaborations among researchers worldwide.

1. Preparation

1. Assemble the semi-degreased whale skeleton.
2. Designate a code for each piece of bone. The code will be used in the photoshoot, 3D model generation, and 3D printing.

2. Photogrammetry

1. Camera and tripod settings
   1. Use a standard lens with focal length of 24-70 mm, diameter of 77-82 mm, and f number of 2.8 L. Avoid wide-angle lenses. Use a compatible tripod with adjustable height of 40-150 cm.
   2. Set the shutter at 1/25 to 1/30, depending on lighting condition.
   3. Set the aperture at f11 to f13. Ensure that the background is clearly captured.
   4. Set ISO to auto. Ensure that the value does not exceed 1600.
   5. Set the tripod at approximately 20-40 cm away from the specimen.
   6. For horizontal elevation, adjust the tripod height so that the camera is horizontal to the specimen (Figure 1).
   7. For superior elevation, adjust the tripod height so that the camera is tilting down 45° above the specimen.
   8. For inferior elevation, adjust the tripod height so that the camera is tilting up 45° below the specimen.
2. Shooting the vertebrae
   1. Prepare a clear tabletop in room A. The dorsal part of the bone will be scanned here.
   2. Position the bone on the tabletop. Use a ring light for bones with residual oil and which appear darker. Optional: place aluminum foil underneath for better light reflection.
   3. Place 2-3 marker cards at known distances as scale reference.
   4. Set the tripod to horizontal elevation (step 2.1.6).
   5. Take a photo, then move the camera by 15-20° circularly around the specimen. Repeat until a 360° turn is completed.
   6. Set the tripod to superior elevation (step 2.1.7). Repeat the  photoshoot (step 2.2.5).
   7. Prepare a clear tabletop in room B. The ventral part of the bone will be scanned here.
   8. Position the bone upside down on the tabletop. Repeat the photoshoot (steps 2.2.4 to 2.2.6).
      NOTE: Room A and B should not have any common objects except for the specimen. The background should remain unchanged to prevent confusion during post-processing. A fixed focal length and distance from the object throughout the photoshoot should be maintained. For scale calculation, at least 2 marker cards must be captured in a single photo. At least 5 photos with marker cards should be captured for each specimen. For specimens with unique features (e.g. foramina) 2-3 extra close-up photos should be captured. All of these also apply to steps 2.3 and 2.4.
3. Shooting the large bones (skull, mandibles, ribs, scapulae, etc.)
   1. Prepare transparent supporters (shelves or boxes) and put the bone on the supporters.
   2. Set the tripod to horizontal elevation (step 2.1.6).
   3. Take a photo, then move the camera by 15-20° circularly around the specimen. Repeat until a 360° turn is completed.
   4. Set the tripod to superior elevation (step 2.1.7). Repeat the photoshoot (step 2.3.3).
   5. Set the tripod to inferior elevation (step 2.1.8). Repeat the photoshoot (step 2.3.3).
4. Shooting the small bones (phalanges, chevrons, etc.)
   1. Prepare a turntable covered by aluminum foil, on top of it place a foam cube with a few toothpicks pointing upward as support for the specimen.
   2. Put the bone on top of the toothpicks so that its ventral part is visible from underneath.
   3. Set the tripod to horizontal elevation (step 2.1.6).
   4. Take a photo, then rotate the turntable by 15-20°. Repeat until a 360° turn is completed.
   5. Set the tripod to superior elevation (step 2.1.7). Repeat the photoshoot (step 2.4.4).
   6. Set the tripod to inferior elevation (step 2.1.8). Repeat the photoshoot (step 2.4.4).

3. Data processing by modeling software (see Table of Materials)

1. Creating a sparse point cloud
   1. Go to Workflow menu, select Add Chunk and Add Photos. Select all photos of a single specimen and click Open.
   2. Go to Workflow menu, select Align Photos and click OK. This step will take some time.
   3. Go to Model menu and select Gradual Selection.
   4. Select Reconstruction Uncertainty, set value to 10 and click OK. Press [DEL] to remove selected points.
   5. Go to Tools menu, select Optimize cameras. Check Adaptive camera model fitting and click OK.
   6. Repeat step 3.1.3, select Reprojection Error, set value below 0.5 and click OK. Press [DEL] to remove selected points.
   7. Repeat step 3.1.3, select Projection Accuracy, set value below 10 and click OK. Press [DEL] to remove selected points.
   8. Rotate the point cloud and delete unwanted points with the Free-form selection tool and [DEL].
2. Cleaning the sparse cloud
   1. Go to Workflow menu, select Build Mesh.
   2. Select Sparse cloud as Source, untick Calculate vertex colors and click OK.
   3. Go to File menu, select Import – Import Masks.
   4. Select From Model as Method, Replacement as Operation, apply to All cameras and click OK.
3. Building a dense point cloud
   1. Go to Workflow menu, select Build Dense Cloud.
   2. Select the Quality (Medium or High), uncheck Calculate point colors, and click OK. This step will take some time.
   3. Rotate the point cloud and delete unwanted points with the Free-form selection tool if necessary.
4. Cleaning the dense cloud
   1. Go to Workflow menu, select Build Mesh.
   2. Select Dense cloud as the source and click OK. Wait for it to load and save the project.
5. Scaling the model
   1. In the Photos panel, double-click a photo with marker cards. Zoom-in and right-click at the center of a marker and select Add Marker. Repeat for other markers on the photo.
   2. Repeat for all photos with marker cards. For previously added markers, right-click, select Place Marker and choose from the list
   3. Adjust the position of markers by holding left-click.
   4. In the Workspace panel, under Chunk – Marker, select a pair of markers with known distance by holding [CTRL]. Right-click and select Create Scale Bar. Repeat for all pairs.
   5. In the Workspace panel, under Chunk – Scale Bars, select the pair of markers.
   6. In the Reference panel, enter the Distance in meter. Repeat for all pairs.
   7. Save the project.

4. 3D printing of the skeleton

1. Printing the model
   1. Export .STL files to a 3D printing software (see Table of Materials).
   2. Move and rotate the object on the platform if necessary. Make sure the object is positioned with a large flat surface touching the base. Scale up or down if necessary.
      NOTE: 3D printing is a very time-consuming procedure, and the irregular shape of bones may complicate the process. The flattest surface should be designated as the bottom part of the model so that the printout remains stable throughout printing.
   3. For parts that are too large to be printed (e.g., the mandibles), use the Free Cut function to divide the object into parts and glue the printouts together.
   4. Print the model with PLA filament. Use pass distance of 0.03 mm.
      NOTE: Printing at a smaller pass distance is recommended for models with fine-scale features, which will require a longer printing time.
2. Displaying the skeleton
   1. Match the 3D printed products with the originals by the designated code. Check for any misprint. Reprint if necessary. Assemble the skeleton for display.

In this study, 166 pieces of bone were scanned individually and the 1 mm resolution 3D models were saved in .STL format. The stereolithography format records surface geometry of 3D objects without color or texture, which is common for 3D printing. The complete 3D model of the Omura’s whale skeleton was uploaded online for public access (https://www.cityu.edu.hk/cvmls/omura). The skull was printed with a professional 3D printer due to its large size. The rest of the skeleton was printed with an in-house 3D printer (Figure 2, see Table of Materials). The mandibles were printed in sections. All the whale bones were matched and compared with the 3D printed products (Figure 3). The final product was largely analogous to the whale specimen and was suitable for general research, exhibition, and educational purposes (Figure 4). However, 2 diagnostic features (i.e., foramina on the parieto-squamosal suture and an indented groove on the frontal at the posterior end of the lateral edge) were not satisfactorily reproduced, which hindered the product from being used for detailed anatomical research.

Figure 1: Setting of camera angles for photogrammetry. To acquire photos from different angles, the camera tripod is set to different heights: horizontal to, 45° tilting down from above, and 45° tilting up from below the specimen, keeping a constant distance of approximately 20-40 cm. Please click here to view a larger version of this figure.

Figure 2: 3D printing of the whale skeleton. A piece of vertebra (Ce1) was printed with PLA filaments using an in-house 3D printer. Supporting structures that were not part of the bone model were subsequently removed in the final product. Please click here to view a larger version of this figure.

Figure 3: Matching pairs of whale bones and the 3D print products. (A) ribs, (B) sternal ribs, and  (C) chevron bones. The bones and the half-sized 3D print products were matched with their designated codes and checked for any misprint. Please click here to view a larger version of this figure.

Figure 4: The complete 3D printed Omura’s whale skeleton. It was exhibited at (A) the City University of Hong Kong and (B) Hong Kong SciFest 2019 in the Hong Kong Science Museum. Please click here to view a larger version of this figure.

Skeletons for exhibition should be oil-free and odorless. Whale bones are notoriously oily, which makes their preparation exceptionally challenging. A juvenile blue whale stranded in 1998 had its skeletal remains exhibited in the New Bedford Whaling Museum, Massachusetts. Despite tremendous treatment by professionals, the bones remained yellowish with an unpleasant odor, and has been oozing oil continuously for over 20 years²¹. With little experience in handling whales in Hong Kong, great effort was spent on degreasing the Omura’s whale specimen, yet the result was unsatisfactory. Composting with horse or elephant manure for a prolonged period was suggested to further remove the oil, yet without cautious temperature control the bones could easily be ruined. Even after degreasing, the mineral composition of bones is prone to biodeterioration by various airborne bacteria and fungi¹⁰. It was therefore decided to have the bone morphology digitally documented by a surface documentation method and to reproduce the skeleton with a durable material.

The data acquisition time for photogrammetry depends on the shape and complexity of the object. In this study, smaller bones like phalanges took about 7 minutes, ribs took about 15 minutes, larger bones like vertebra took about 45 minutes. Data acquisition of the entire skeleton was completed in 4 weeks. The limitation is that photogrammetry does not produce the model in real time as 3D surface scanning does. Only after the photos are imported and processed the operator can realize whether they are sufficient to generate a complete 3D model. If a section is inadequately recorded, the photoshoot should be repeated under the same condition to supplement the missing geometric data, since difference in ambient lighting may cause variation in model quality. The post-processing time for photogrammetry is also longer than that of 3D surface scanning²². It took about 14 weeks to produce the digital model of the entire Omura’s whale skeleton. Overall, the 3D documentation part was completed in 18 weeks. Noise reduction hole filling is necessary to fix errors in the meshes, commonly as non-manifold edges or vertices²³. A powerful computer is preferred for processing large number of photographs.

Printing of the full skeleton took about 10 weeks and costed approximately US$6500. The PLA filaments showed a realistic stereoscopic sensation comparable with the original. The product is lightweight, resistant to ambient moisture and microbial growth, mechanically stable, and relatively inexpensive. PLA is non-toxic and environmental-friendly, as it is derived from crops like corn and sugarcane. The only concern with PLA is avoiding exposure to sunlight or temperatures over 60 °C, which can make the polymer brittle and break. In total, approximately 28 weeks were needed to complete the photogrammetric 3D modeling and printing of this Omura’s whale skeleton.

In this study, at least 4 unique diagnostic features were observed on the original whale skull. The outline of the rostral portion that bowed out more than in other whales, and the caudal condylar facet that articulated with Ce1, were satisfactorily presented in the model. The foramina on the parieto-squamosal suture on the posterior wall of the temporal fossa, and the indented groove on the frontal at the posterior end of the lateral edge, both of which are diagnostic for B. omurai⁹, were not as clear in the 3D printed product. Since photogrammetry largely relies on photo quality and angle of view, a routine photoshoot may omit local details with recessed contouring. Prior to the photoshoot, operators should be informed of anatomical regions of particular interest, so that extra photos are captured to yield 3D models with extra precision. Photogrammetry is limited in capturing “shapeless” structures, for instance, foramina and sutures that do not show any 3D delineation²⁴. Restoration of such topographies may require post-editing of the 3D models. Alternatively, X-ray, CT and 3D surface scanning can be used in conjugation to document subtle morphologies, or even internal structures with high resolution^{14,15,16,25,26}. Radiographs can be digitally combined with 3D models from photogrammetry to form “pseudo-CT” images, while 3D surface scanning can supplement information on color with high fidelity²⁷.

Because cetacean strandings are highly opportunistic, only a limited number of deceased animals were retrieved properly, not to mention the tedious effort invested in carcass treatment. A meticulously preserved cetacean skeleton is treasurable and requires careful maintenance. Digitalization of morphological data can make the best use of specimens. The workflow presented in this work from carcass treatment, 3D documentation to 3D printing is revolutionary for both research and education collaborations.

The authors have nothing to disclose.

The authors would like to thank the Agriculture, Fisheries, and Conservation Department and Marine Region of the Hong Kong Police Force of the Hong Kong Special Administrative Region Government for their support in this project. Sincere appreciation is also extended to the staff and students from the City University of Hong Kong for the great effort placed on defleshing and treating the Omura’s whale skeleton. The authors gratefully acknowledge Department of Infectious Diseases and Public Health of the City University of Hong Kong for the financial support on this publication cost. Special thanks to Dr. Maria Jose Robles Malagamba for English editing of this manuscript.