Chocolate Plantation
Laser Scanning and JRC Reconstructor
Overview
The goal of this project was to create a 3D model from data gained through the process of laser scanning. In this process a structure or object is scanned by a laser scanner and photos are taken. In this instance we used a Faro Focus laser scanner which takes photos while it scans the structure. This scanned data is saved in the form of “point clouds” which are clouds of data that together are used to represent a three dimensional structure or object.
For this project we scanned a site known as Chocolate Plantation on Sapelo Island, Georgia. Chocolate plantation was a cotton plantation located on Sapelo Island, though the land was purchased much earlier agricultural activity did not begin until 1793. Lewis Harrington purchased the property and utilized the labor of 68 slaves.¹ The construction of the tabby structures in question did not occur until Swarbreck’s purchase of the land in 1808.² The tabby structure that was scanned was believed to be constructed around this time and served as planter’s residence.³ During the Civil War plantation activities on Sapelo came to a halt. Some slaves remained on the island, but the plantation remained largely abandoned until 1870.⁴ The Green Family resided on the plot well into the 1910s, until the property was purchased as a hunting preserve in 1912 by Howard E. Coffin.⁵
The site scanned was the ruins of the planter's home. The majority of the structure was no longer present besides the back wall and chimney and the perimeter wall.
First Phase: Scanning
The first part of the project was the scanning of the site on Sapelo Island. To do this we used a Faro Focus laser scanner which was brought to the site on Sapelo Island. An initial survey of the site was conducted before scanning, examining the surrounding structures and photographing them. The structure that was scanned was the ruins of a tabby cabin on the south-west side of the property. The majority of it was no longer standing, save for the back wall and chimney which made these ruins in particular a point of interest. The ruins had no support system keeping the back wall from collapsing so it was important to scan this structure now while it was still standing as opposed to later. The other ruins on the property seemed to be in a more stable state, some even had additional support systems added at a later date to keep them from collapsing.
The next step was to set up the Faro Focus laser scanner. It was set on a tripod, first in the center of the ruins to get a 360 degree scan of the interior of the ruins. The scanner does an initial preview scan which takes a few seconds. Then, using the onboard interface, the scan area is cropped down to exclude areas that aren’t of interest. Then, the scanner begins the scanning process. This is an automated process though it takes a few minutes depending on the quality of the scan and the scan area. After the first scan was finished it was moved to another area to scan the ruins from a different perspective. This was done 12 times to capture as much of the structures as possible. It is important to have a significant amount (roughly 50%) of overlap between scans so they can be easily stitched together in the registration phase.
Faro Focus scanner and scanned tabby ruin. Photo by Hew Evans
Second Phase: Processing & Registration
After returning from scanning the next step was to process the raw scan data. The program that was used for this step was FARO Scene. The 12 raw scans were imported into the program where they would be processed. This step was automated and took about an hour. The processed scans could be previewed in the program to make sure that they came out properly. The next steps of the process took place in JRC Reconstructor.
The processed scans from Faro Scene were brought into JRC Reconstructor. In this program they could be viewed in 3D space and edited and registered. The first step was to crop the scans to include only the important information. The scans captured information outside the intended scan area, such as background trees and grass, that could be deleted. Doing this was simple, only requiring a select tool and a delete tool to remove the unnecessary information. This was done for all 12 scans until they only included the intended scan area, the perimeter wall of the ruins and the backwall with the chimney. The next step was manual registration.
Manual registration is a way of “stitching” the scans together. Since each scan is from a different perspective they need to be aligned properly to create a whole 3D model. This was done with a tool in the program called “Manual registration.” This tool placed one scan as a reference grid and the other as the moving grid.
The reference grid was the scan that was referenced and stayed stationery and the moving grid was the scan that moved to align itself with the reference grid. Since this was a manual process, it required a number of overlapping points (a minimum of three) between the scans to be chosen. These points would be used as reference for the program to align the scans by.
This was largely a trial and error process, as it would sometimes result in an improper alignment. After the two scans were aligned properly the process was done for the next scan, using either of the now aligned scans as a reference grid and the unaligned scan as a moving scan. This process was done for all 12 scans.
Lining up the scans through manual pre-registration
Manual pre-registration interface
Afterwards the scans should be lined up in a way that properly represents the scanned object as an accurate 3D model. Sometimes there are small areas that don’t line up properly and have to be aligned using a different tool. This next tool was a automatic registration process which used a similar technique as the first tool but instead of manually picking the points the program did it automatically. This automatic tool is intended for more finite adjustments once the scans are properly aligned. It uses a reference scan and a moving scan and aligns them automatically. This was done with all 12 scans, resulting in a properly aligned 3D model.
The next and final step was converting the model from a point cloud to a mesh. This is done relatively easily. First all twelve individual scans, or point clouds, had to be combined into a singular point cloud. This was done through selecting all twelve scans and combining them into a singular point cloud. Next, the mesh conversion tool was used to create a mesh from the data. This meant creating a solid 3D model instead of a collection of points in space. This process was automatic though required a few minutes to process, depending on the quality of the mesh selected. The output was a 3D model in .obj format that could be used for other purposes and viewed in other programs.
References
1 Devan, L. Richard. “Pieces of Chocolate: Surveying Slave and Planter Life at Chocolate Plantation, Sapelo Island, Georgia.” African Diaspora Archaeology Newsletter 11, no. 2 (2008). Page 4
2 Ibid
3 Devan, L. Richard. “Pieces of Chocolate” African Diaspora Archaeology Newsletter 11, no. 2 (2008). Page 7 - 12
4 Devan, L. Richard. “Pieces of Chocolate” African Diaspora Archaeology Newsletter 11, no. 2 (2008). Page 5
5 Ibid