The Process
Scanning Process
The 3D scanning process started with setting up the scAnt system by installing the required software, connecting the camera and motors, and preparing the scanning area. After creating a new project for our first specimen, we made sure there was enough storage space, since a full scan could take up more than 200GB of images. We then configured the camera settings, including exposure, gain, white balance, and focal length, to match each specimen’s size and color. These settings had to be adjusted every time to make sure we captured each specimen as clearly as possible. We also set up the flash timing to ensure the lighting was even, and enabled post-processing options like image stacking and background masking to make the final images more focused and cleaner.
To carry out the scan, we rotated and moved the specimen and the camera along three axes — X (pitch), Y (rotation), and Z (focus) — by entering minimum, maximum, and step values. This allowed the camera to take a series of overlapping images from different angles and depths. For each angle, multiple photos were taken at different focus levels, which were later combined into one sharp image using a stacking method. This process helped capture the full detail of each specimen and generate a set of images that will later be used in a photogrammetry software for the 3D reconstruction.
Image Processing
After the scanning process, we used a photogrammetry software (3DF Zephyr Lite) to process the images and create a 3D model. The software automatically aligned the images, generated a sparse point cloud, then we generated a dense point cloud, and then created the mesh. We then cleaned up the mesh by removing any unwanted artifacts and filling in holes. The final step was to texture the model using the original images, which added realistic colors and details to the 3D model.
The resulting 3D models were then exported in OBJ and STL, which are commonly used for 3D printing and virtual reality applications. The models were then imported in blender, in order to clear the vertices that correspond to the scanner's base and the specimen's pin and thus isolate the specimen model itself. Lastly, the cleaned mesh was then exported and uploaded on sketchfab, to make the model interactive and we used sketchfab to embed it in our website, so that the users will not have to navigate to sketchfab for every specimen.
This is the Sparse Point Cloud of the mesh, which is a 3D representation made up of a small number of points that mark key features or structures of an object or scene. It’s usually the first result generated in photogrammetry or 3D reconstruction, based on identifying and matching points across multiple images. While it doesn’t capture fine detail, it gives a rough shape or outline and is used as the foundation for creating a dense point cloud or full 3D model later in the process.
This is the Dense Point Cloud of the mesh, which is a detailed 3D representation made up of a large number of points that cover the entire surface of an object or scene. It’s generated after the sparse point cloud and includes much more information, capturing finer textures, shapes, and depth. Each point in a dense cloud represents a part of the surface, often with color data, making it useful for creating accurate 3D meshes and textured models.
This is the final reconstruction of one of our specimens- Vanessa Atalanta. It is the textured mesh that was constructed with the dense point cloud, giving us a high detailed 3D representation of the specimen. The difference from the Dense point cloud is that this is the actual generated model that is exported and uploaded in Sketchfab after some cleaning in Blender
