On March 2nd 2013 the Zodiac arrived in South Lake Union and tied up next to the Center for Wooden Boats to wait for the dry dock to be ready. Our friend Ross from Toronto arrived the day before and brought with him a very expensive piece of equipment, a Faro Focus 3D Sanner which will set you back by ~$40,000. This scanner has a range up to 393 feet (120m) and an accuracy down to 1/16 of an inch (1-2mm) which means we will be able to now get very precise measurements of the Zodiac. The funny thing is that this level of accuracy on a wooden boat doesn’t make a lot of sense since measurements actually change constantly. Wood is a very flexible organic material that twists and bends based on temperature and forces applied. So for example the measurement of the boats length in dry dock is very likely different than when it floats in water. The trick question is how much of a difference can we register. We’ll get back to that later.
On Sunday I met Ross on the Zodiac and we got busy right away scanning the boat. Ross took two scans from the dock which gave us some initial point cloud data to play and experiment with. We then moved on to scan the entire inside of the boat from the chain locker in the bow to the lazarette in the stern. We even placed the scanner into the engine room, the bilge with the propeller shaft below the main lounge and even the very tight forward bilge. In total we created 18 scans of the inside of the Zodiac that resulted in about 600-800 million scan points or roughly 40 giga bytes of uncompressed files on the hard drive. Or in other words for each square inch of surface we have about 2,500 points each with X, Y and Z coordinates and an RGB color value. Now that is a lot of data that needs to be processed and I quickly learned that my desktop computer with a fast quad-core 64bit processor and 12gb RAM will be pushed to its limits. I ended up buying 16gb more RAM memory and learned that this high data density isn’t really necessary to what we are trying to do.
On Tuesday we finally got into dry dock and you can see the entire process on a time laps video I created with my GoPro camera. The video is posted to YouTube and compresses the 2 hours process into 44 seconds. Once the Zodiac was dry we got immediately to work to setup the scanning process. The picture below is a complete point cloud of the entire boat inside the dry dock. This point cloud was assembled from 29 individual scans along the bottom of the hull and from the top of the dry dock walls and even the Zodiacs deck itself. The scanner always does a 360 degree scan so it picks up everything around it including the dry dock, the crane and buildings on shore and even other boats around us. The mystical colors come from the fact that Ross started scanning at 5pm and worked way into the night with only the top mast light illuminating the scenery.
Once all the scanning was completed we had almost 2 billion scan points which used uncompressed about 80gb of space on the hard drive. Ross and myself got busy processing the data because we wanted to show some preliminary results to Tim before Ross flew back to Toronto. We wanted to provide a number by which the port and starboard side of the hull is different. During Restoration they found quite a bit of damage on the port side of the hull due to the 40 year use as a pilot boat.
One of the first steps was to align each scan for which we used 12 special 6″ spheres with magnets that we placed along side the walls of the dry dock. Each individual scan needs to pickup at least 3 of the spheres which the scan alignment software Faro Scene auto-detects and highlights. So all 29 exterior scans included at least 3 of the spheres for which the software calculated precise 3D coordinates. These know reference coordinates are now used to align each scan down to 1/16 of an inch accuracy as long as the sun is not shining. Well what does the sun has to do with this you might ask. The dry dock is made out of steel and is about 49 feet wide from wall to wall. If the sun hits the walls they heat up and expand which now might change the width to 49 feet and one inch. It might also change the length by an inch after which the reference spheres are no longer at the precise positions for each scan which would result in the scan alignment being off by a couple if inches. However, we got lucky and had an overcast day with fairly constant temperature and very little wind which can have a similar effect as the the sun.
After we aligned all 29 exterior scans we started to clean up the scans and delete all scan points that are not part of the boat. We removed all the dry dock walls, buildings and the crane, dry dock floor and the supports that keep the Zodiac in position. Then we removed all the lines and the ladder we use to get on the boat in dry dock and reviewed all hull surfaces for other scan data that is not part of the hull. We also removed all deck structures and the rigging since for this initial analysis we only need the hull surfaces for each side. This process took quite a few hours and you can think of it like “Photoshop”ing a picture. The only difference is that we do it in 3 dimensions. Below you see the result of that effort.
Once we had fairly cleaned up hull surfaces we actually split the point cloud in half right through the center plane of the boat. We used for this some free open source point cloud management software called “CloudCompare” which is actually pretty amazing. It allows you to work with very large point clouds as long as your computer has enough RAM memory. You can reduce the data density by re-sampling the point cloud to have 10% of the original size. So our 2,500 points per square inch were reduced to 250 points. Then we mirrored the starboard hull point cloud right on top of the port side one.
The software now processed the spacial difference between each hull surface and calculated for each point pair precisely how far are they apart. The result can now be visualized by a color scale that shows how fare certain areas of the hull are apart. As you adjust the color scale (the scale in the pictures is in meters. So 0.3m equals roughly 12 inches) you can start to see the actual difference between each surface and the expected port side indentation of the hull becomes very visible and quantifiable. We can see a maximum difference between the hull surfaces of up to 3 inches which is significant but not a surprise nor an issue regarding the hulls structural integrity. It just shows how flexible the wood hull of a large vessel such as the Zodiac actually is.
Next up is further data clean up and interior alignment which is much more difficult because we don’t have reference spheres and I need to rely on surfaces that overlap with each scan. After that I will start to actually reconstruct the hull surface by turning the point cloud into a mesh and then ultimately into a CAD model.