Recent Changes

Wednesday, January 7

  1. page home edited ... Building 3D models with Digital Photographs A guide to Small-Craft Monitoring, Preservation, …
    Building 3D models with Digital Photographs
    A guide to Small-Craft Monitoring, Preservation, and Documentation
    Funded by the Institute of Museum and Library Services
    (This Wiki is still under construction and lacks formatted photographs)
    Kyle HunterHunter, The Center for Wooden Boats
    Contributors: Jonathan Taggart, Jack Becker, David Cockey, Kathrine Cockey, Todd Croteau, Dana Lockett, Don Rothwell, Eric Hervol, Nat Howe
    {Cover Photo 1.JPG}
    Software: Agisoft Photoscan and Cloud Compare
    Equipment Cost (one time): $3,500 (includes digital camera and laptop which can be used for multiple purposes)
    This screen shot captures a comparison (middle) the between an Agisoft Photoscan point cloud (left) and a LiDAR point cloud (right). The pole depicted in the center is the comparison of clouds aligned using Cloud Compare software. The yellow, orange, and red coloring that you see represents minor discrepancies between the two models. The biggest differences are in the chin of the face of the figure holding the canoe and the canoe itself. The difference in the chin is due to the shadow of the chin in the photographs used to build the model in Agisoft Photoscan. The difference in the edges of the canoe that the figure is holding stems from the difficulty of the Agisoft Photoscan software to process edges in detail.
    Note the “garbled” section at the top of the model built in Agisoft Photoscan. This is due to the silhouetting that occurred against the sky, darkening the object, and the failure of the software to align the photos properly. This is also reflected in the screen shot below, where many points are missing. This could be remedied by taking additional photos and adding them into the software and ultimately the model.
    It is apparent that both technologies are capable of comparable results, as long as the Agisoft Photoscan process was thoroughly done. We would have better results in our Agisoft Photoscan model if more pictures were taken in areas that had pointed, thin edges or deep shadows.
    Deciding which method might work better for this type of tall object would depend on the circumstances. Both models needed more information as you go further up the pole.
    It is apparent that both technologies are capable of comparable results, as long as adequate photos are taken for building the 3D model.
    Nordic Spirit is a large, Viking-style ship built in Northern Norway approximately 200 years ago. It is currently owned by the Nordic Heritage Museum in Seattle, WA, and is on permanent display under a covered outdoor section in front of the museum. Because of the age of the vessel and the challenges of preserving it, it is prudent to make a good base model to monitor its stability.
    Nordic Spirit was scanned using a stationary LiDAR laser scanner, an instrument capable of capturing immeasurable detail. In this case, the vessel was supported by several obstructions that left significant “holes” of information in the model. Due to the nature of the scanner and limitations in mobility, it is not possible for the scanner to read obscured areas.

    A view of the main obstruction from the support system.
    A model of the boat was also built in Agisoft Photoscan, using the same process outlined in previous studies of this document. While no technique is able to penetrate obstructed sections of the object, a camera is small enough to work around in the majority of tight spaces. A more complete model was created using this method. This model is a good base to compare future models to in order to monitor the stability. It also serves as an illustration of the versatility of the Photoscan method.
    Comparison of Total Station to Photoscan- Salish Canoe
    In 2008 4Culture gave The Center for Wooden Boats a grant to document two vessels owned by CWB. The vessel of this particular study is a unique 15’ 6” Salish Dugout Canoe. In 2009 Todd Croteau of the National Park Service HABS/HAER program came out to help CWB document this vessel, of which resulted in an Historic American Engineering Record (HAER) being filed with the Library of Congress.
    Equipment Cost (one time): $3,500 (includes digital camera and laptop which can be used for multiple purposes)
    Training individuals with these methods can be tricky. As with any process, a new skillset might be required. In each case, there is hardware (total station vs. camera) and software unique to each process that must be mastered. The learning abilities of each individual will determine the speed and accuracy at which these methods can be used. They are similar in complexity and therefore cannot be argued that one method is easier to use than the other. It was gathered through experience and discussions that each method is comparable to labor spent on each result.
    actual object. {999.JPG}
    Here the lines after the author “faired” them, as they were derived from the Agisoft Photoscan, Rhino 3D, and AutoCAD process. Those “faired” lines (red) are overlaid onto the lines derived from the Total Station process (black). The reader can see that they are close in proximity. It must also be noted that the 3D model built of the Salish canoe was one of the first done in the project, and therefore of a rather low quality. As the user learns the process, results are dramatically improved.
    Here the lines derived from the Agisoft Photoscan, Rhino 3D, and AutoCAD process (in red) are overlaid onto the lines derived from the Total Station process.
    Photographs for the process of photoscan can be obtained at any time, and archived for the future. Even if a model were to never be built, hundreds of detailed photos of the object would exist in case it was ever altered, destroyed, or lost.
    If a small institution chooses to invest in Method 2, half of the cost is in the camera and computer necessary to capture and process the photos. These items might already be owned and can be used for multiple purposes, thereby potentially reducing the impact on the financial constraints of the organization.
    Overall Conclusions- start taking photographs now
    The intended result of this project was to develop a cost effective, simple method to monitor the condition of irregularly shaped large objects, in order to best preserve the cultural heritage specific to regions and peoples. Though the preservation of these objects is the ultimate goal, it is inherently impossible to prevent the deterioration and ultimate destruction of organic materials. Luckily, the methods studied go well beyond monitoring an objects stability and reach deep into the preservation of culture.
    (view changes)
    7:07 am
  2. file image035.png uploaded
    7:06 am
  3. file image035.png uploaded
    7:04 am
  4. file image034.jpg uploaded
    7:04 am
  5. file image033.jpg uploaded
    7:04 am
  6. page home edited ... Note: This section contains terminology specific to Rhino3D, so a basic understanding of this …
    Note: This section contains terminology specific to Rhino3D, so a basic understanding of this program will be necessary in order to fully comprehend it.
    While lines plans with table of offsets and detailed construction drawings are very helful for builders to take on a new construction project of an original classic, it isn’t always necessary to create them in order to preserve all of information. Many boats do have their original documentation, and of those that do, it might be more important to document the specific object for cultural purposes; i.e. the rowboat that took George Washington across the Delaware River vs. other boats built by the same builder of the boat that President Washington used. Creating a study such as this has potential to reach a wider audience using a more general interpretation in order to grasp the details of an object that can’t be touched or easily seen.
    An orthographic projection is “projection of a single view of an object (i.e. a perpendicular view of the object) onto a drawing surface in which the lines of projection are perpendicular to the drawing surface” Merriam-Webster, This means that all perspective distortion has been eliminated from views such as the top, front and right side views. However, it does not apply to the rotated perspective views (see green arrow above).
    The above montage is comprised of drawings and orthographic photos imposed over those drawings. This was created to show how the textured (rendered) orthographic projections of the scaled 3D model can be used in conjunction with lines drawings. One of the advantages of this type of presentation is the amount of cultural information included in one page of documentation. As the drawings for the Mukilteo boat in the previous study were created by the user’s judgment to record the lines in what makes sense in terms of construction, this can lead to a compromise in detail of the original object. There are often variations in the actual object to its original state, such as off center frame placement or repairs that have altered the object from its original construction, all of which might be erased when creating a lines plan and construction drawings. The variations that are evident in the Davis boat (above) is a good example of how this method of record can capture the object as-is, while still allowing the reader to use a divider and scale to garner qualitative measurements from it. Therefore, these textured orthographic projections are a more accurate form of documentation closer to the original object. Additionally paint colors and even the shadows of the boats numbers can be seen, which would be eliminated from drawings, though possibly noted in the report and definitely included in any accompanying photographs.
    The Layouts chosen in the montage above and the level of detail entirely depends on the project goals determined beforehand. This study is meant as a primer to illustrate some possibilities of what can be accomplished using this technology.
    Monitoring the change in shape of an object using Cloud Compare
    cost at [[file:/C:\Users\keelb_000\Documents\IMLS%20project\|]]
    Large object collections are often stored in less than ideal conditions so it is incumbent on collection managers to monitor the condition of those objects carefully. This presents its own set of challenges. By the time a change in shape or size of a large object is perceptible by the eye, significant, irreversible damage has already occurred. To accelerate the process to get to a conclusion on whether these software programs were capable of recognizing the change in the shape of a hull, CWB used a derelict recreational fishing boat built in 1929 to physically manipulate that very change.
    A 3D model was built using the same method explained with the Mukilteo boat above. To simulate the shape change that might occur over time, a lead weight was cantilevered out over the gunnels of the boat, in opposing directions, giving the hull a definite and perceptible twist. Then, mimicking elapsed time of a lengthy period, as might occur with an improperly supported boat, we created a second 3D model. We repeated the photographic process to minimize any possible errors arising from a change in camera settings or photographer, and built models using the same parameters in Photoscan.
    Exporting the models from Agisoft Photoscan as a standard .ply file and uploading them each into Cloud Compare, we were able to align the models to the points most common to each model. Cloud Compare provides the user with a heat map type function to illustrate the differences in the shape of the 3D model, thereby highlighting the differences in our physical hulls. The red coloring illustrates the most extreme differences between the two models, while the areas without coloring represent sections of the boat unaffected by movement.
    In order to qualify the actual distance of movement into relative units of measurement, the model needs to be scaled in Rhino3d prior to uploading into Cloud Compare. More overall research is needed to determine the level of accuracy using cloud compare, as the project was only able to test simulation of a shape change of one object. However, based on other studies in this project, the accuracy level is estimated to be roughly 1/8inch error over 20 feet.
    If highly accurate measurements are needed before further research is conducted using the baselines of other models in this project, the total station survey is a proven method to measure that movement. The Vhasa Museum ( has done an excellent job of geodetic measuring using total station survey that produces results with an accuracy of less than 1mm.
    The Center for Wooden Boats cares for a fleet of eight Blanchard Junior Knockabouts (BJK). These boats were built on Lake Union in Seattle, WA during 1933 and 1947 to give customers access to an affordable “daysailer” during the depression. The design was a shortened version of the Senior Knockabout, and the story goes that building the first few BJK’s too many planks were broken during construction due a strong curve at the bow of the boat. In order to prevent this from happening further, the design was elongated slightly to lessen the curve. Of the eight that CWB owns, one is the shorter version at 19’ LOA compared to 19’8”. While the length is easily distinguishable to the eye, the complex curvature in the shape of the bow is more abstract to comprehend and ultimately compare by hang measurement.
    Building a model of each vessel in Photoscan, we exported them and uploaded them the same way we did the two models of the Reinell fishing boat. The white hull (top) is the shorter, earlier version and the stripped hull (below) is the latter, elongated version.
    Cloud Compare aligns the majority of the points together automatically, creating a model that reveals the changes as layers are peeled away. In the photo below, the green hull represents the elongated version. Illustrated is the drastic “bump” as the models combine at the stem of the bow. Further towards the after section of the BJK the difference in the height of the boat is revealed. The elongated hull is narrow and sleek whereas the shorter hull is stubby and thick.
    This provides another view of the differences, especially in the length and height of each hull.{992.JPG}
    Layers of objects can be peeled away to reveal an section that could be of interest
    (view changes)
    6:58 am
  7. page home edited ... A guide to Small-Craft Monitoring, Preservation, and Documentation Funded by the Institute of…
    A guide to Small-Craft Monitoring, Preservation, and Documentation
    Funded by the Institute of Museum and Library Services
    formatted photographs)
    Author: Kyle Hunter
    Contributors: Jonathan Taggart, Jack Becker, David Cockey, Kathrine Cockey, Todd Croteau, Dana Lockett, Don Rothwell, Eric Hervol, Nat Howe
    It is also recommend to take measurements (and photos of those measurements) of the object, so that they can be factored into the model at a later date. For example, one can take a specific measurement from the tip of the stem to the tip of the stern. While these photos may not be useful for the model construction, they will always remain as notation included with the photo documentation. That way they can’t easily be lost as they could be written on a piece of paper. One can then manually pick out specific points in the model and set the scale based off of those specific measurements.
    Note: that the scale cannot be added to the 3D model in the standard edition of Agisoft Photoscan, though that feature is available in the professional version. Instead, the project team is used Rhino 3D for that purpose, which will be discussed in more detail in that section.
    Sample of scaled photos including a tape measure below.
    Note: The 3D model will be only as complete as the photo sets that are taken. Take the time for proper set up and quality photographs.
    The lighting in the warehouse was overhead fluorescent, suspended about thirty feet above the object. On average, half of the fixtures were functional. A Canon G12 camera on variable settings, both automatic and manual, was used to capture the photos. The photos were taken in an orderly and consistent way, to make them easier to keep track of while processing. To capture the hull, roughly 30 photos were taken clockwise around the boat at eye level, keeping a consistent height locked on the tripod and a roughly consistent distance from the object regulated only by sight. The next set was raised a couple of feet to get a slight downward angle, and shot clockwise in a similar fashion.
    To get the bottom of the hull, which was especially dark as it was not lit well, a manual setting with an open aperture and slow shutter speed allowed the photos to show the details of the surface. If a photo is too dark, the details cannot be “read” by the program. The camera was affixed to a self-built wooden contraption (floor mount) to which an adjustable ball joint camera mount was fastened. This allowed the camera to be locked into a position so that the lens was facing up toward the bottom of the boat (and ceiling). With a remote control shutter release, the contraption was pushed along snapping over-lapping photos, back and forth under the boat until the entire bottom was captured.
    Tips for photography for use in Agisoft Photoscan
    Photos should be taken anywhere from 4’ to 8’ ft. away from the object, depending on the size. As the object increases in size, generally so should your distance. However, a closer range will produce more detailed results
    Combining photos from different sets
    Twilight is a NW built 36’ fishing trawler built in Seattle in 1933 by Harold Hansen. A once common double ender of the era, she was acquired by Northwest Seaport, a partner organization for this grant, in 2000. In December of 2013, while floating in the water, the team comprised of project partners listed in this grant took a complete set of digital photographs from various vantage points around the vessel. Twilight was moved repeatedly throughout the photo shoot, and photographs were taken from other floating boats and from the pier. A tripod was not used for all shots.
    The resulting model generated in Photoscan was remarkable. Two other common ways of capturing enough data for generating a 3D model, stationary LiDAR (laser scanning) and Total Station survey, are not capable of generating a model of an object in motion. Note that because the boat was in the water, the model lacks any information below the waterline.
    Later, in July of 2014, Twilight was hauled out and trucked up to Pt. Townsend, WA for storage out of the water. Nat Howe, present for the original photo shoot, took another set of photos after she was blocked up. The model that we created of this latter shoot was “merged” in Photoscan with the model from the former, resulting in a model of the boat in its entirety. NW Seaport has a base model of Twilight to monitor her dimensional stability by comparing it to later models built from future photo shoots.
    Photos can be added and processed at a later date to make a more complete model
    Note: As these objects are not perfectly symmetrical, and construction is not always square and fair, it is at this stage that lines plans become interpretations of the actual object. If any artistic liberties or judgments are made in this interpretation, it is valuable to make note of that in any final reports.
    It is in Rhino3D that the traditional lines of the boat are added, and the model in divided by those lines into the sections that will represent the boat in the drawings. The green lines are Station Lines, red lines are Waterlines, and the blue lines are Buttock lines. It is worth noting that proper orientation of the model is critical before creating these lines. The model should be level and plumb as much as possible. The sawhorses and blocking are integral to the boat. Removing them at this stage would leave “holes” in the model. Additional masking of the photos in Agisoft Photoscan prior to creating the model may have reduced or eliminated some of these unwanted components. Supports and stabilization methods should be kept as minimal as possible to increase the view of the surface of the object, but small discrepancies such as these are relatively easy to correct in Rhino3D.
    Refining the lines in Rhino
    The lines follow the surface of the model, including the sawhorses. Agisoft Photoscan can produce reasonable lines of complex curves, yet edges and corners are generally not as distinct. However, higher quality models usually have better representations of corners and edges.
    Additional editing can be performed in Rhino3D to provide missing data, such as plotting the theoretical points along the edge of the transom. This gives the user the ability to correct for inaccuracies created by Agisoft Photoscan, as its tendency to round off sharp edges or to distort certain geometries can make it difficult to automatically extract crisp lines. This is also the case with the edge of the deck and the rub rail, which may have two relatively sharp edges in close proximity and also obscures the true sheer line. An example of this “fairing” technique is shown below. The software user extrapolated the edge of the transom using the Red lines.
    The screenshot above represents the port side of the transom. Note the roughness of the waterlines from the model. The smooth lines represent a faired interpolation of those lines, which were then extended to define their intersection points. The series of points derived in this manner define the theoretical edge of the transom and are included in the Table of Offsets on the Lines Drawing. (For scale reference, the actual distance between waterlines is 3”, so the roughness of the lines and the margin of error appears quite small over most of the model.
    Similarly, the diagram above is an example of how the sheer and rabbet points can be derived at each station section. The surfaces of the hull and deck are very consistent until approaching the deck edge, which is severely distorted due to the presence of a rub rail. Drawing faired lines through the “good” portions and then extending them to their logical intersection produces the desired point on the Sheer line. In the case of the rabbet, the faired line of the hull is simply extended to the half-breadth plane of the keel.
    Table of Offsets
    the hull. These
    are the
    As good as Rhino is for 3D modeling, it isn’t so good for 2D drawing preparation, so these lines were exported into AutoCAD.
    The Lines Drawing documents the size and shape of the hull, such that a new hull can be built to the same shape if desired. In actual practice, when building a boat, the Lines are drawn, or Lofted, full size in order to correct small errors in the offsets and to provide full size patterns for molds, transom, stem, and certain other components.
    The Construction Drawing documents the parts and pieces of the boat and how they are assembled. The Photoscan model is superimposed on the drawing to aid in locating the various components of the hull (frames, engine stringers, thwarts, etc.), but the model is not accurate enough in itself. Manual measurements or scaled photos of each component are necessary for an extremely accurate drawing.
    Direct tracings of oddly shaped individual components like the one below can also be scanned and traced in AutoCAD.
    Agisoft Photoscan is an excellent tool for capturing an accurate scale of 3D shapes of objects, within an 1/8 of an inch over 20 feet
    (view changes)
    6:51 am

Tuesday, January 6

  1. file 9992.JPG uploaded
    2:42 pm
  2. file 9991.JPG uploaded
    2:42 pm
  3. file 999.JPG uploaded
    2:42 pm