Process Document

May 21, 2001.  Visit to Exell Helium Plant- Tour/History.

  Before going to scan we toured the Exell Helium Plant to give us an insight into the historic processes of refining helium that had gone on there for so many years. We toured all the areas that we planned on scanning- such as the Amine Glycol Solution, Carbon Dioxide Building, and the Separation Building. During the tour we discussed the level of detail that would be required in our final drawings. Without this trip we wouldn't have had any idea where to start when it came to scanning and we wouldn't have had any basis for our planning procedures.

  For many of us this was our first scanning experience in the field.  The afternoon before leaving we took the time to meet with Tim Woodruff from Cyra so that we could set up a plan of execution for the next day. During this meeting we sketched the approximate layout of the Amine Glycol Solution area and discussed the best possible scanner locations for maximum time and work efficiency. The first order of business in the field, after making a plan, is setting up the registration targets and spheres. This involved positioning a person at each of the proposed scanner locations while the others walked the targets to locations visible by all the scanner locations possible. This took quite a bit of time but it was beneficial because if a critical target or sphere wasn't visible or was left out it would have taken two or three more hours back at the office on the computer to find a suitable “replacement” and to model it in several different views, as would be needed. We found that good planning in the field saved huge amounts of time back in the office, this would be proved true for us many times. Once we had as many targets set up as possible, we started our first scan.  The laptop should be booted up and interfaced with the scanner before turning the scanner on and once the scanner displays the “ready” light- it is time to start. Next we connected to the scanner and took a picture utilizing the digital camera within the scanner head. In the bright sunlight adjusting the exposure within the Navigation window was not always enough, so we found that placing a pair of non-prescription sunglasses over the digital camera lens helped immensely to see the image displayed on the laptop. It was then that we learned that a lens filter large enough to cover the digital camera area would be necessary during future outdoor scanning. Once we were pleased with the view we had taken, we set up the scanner to start scanning. A very important note we found was as soon as the scan button was selected opening the viewer immediately was mandatory. This allows us to make certain that the scanner was scanning properly, so at the end of the scan we were able to acquire the tie points and spheres. Opening the viewer at this time also ensured that the Scan World would be properly connected to the scanning data once the scan was finished. While the scanner was scanning we took field notes recording Scan World number, operator, scene range, sample spacing, number of points, and any additional important observations. We then auto-acquired the tie points and spheres, but always had to make sure the scanner found all the targets. It took time to hand pick the targets if they weren't auto-acquired because they are often times difficult to see depending on how complex the area of scanning was. In any case, it was better to find them at this point and to get a fine scan on them rather than to try to find them later when we would only have a few data points. The rest of the scanning process involves repeating the same steps; we just moved or rotated the scanner and took the next scans. A real time saver involves what Tim Woodruff calls a “hot move”. This consists of moving the scanner, laptop, and battery while they are all still connected and running. This saved an immense amount of time because we could immediately start scanning our next location, instead of spending seemingly useless time waiting for the scanner to warm up (which was necessary every time the scanner was turned off and restarted). We discovered "hot moving" to be easy  and extremely efficient, especially when you have three people moving short distances, though we found that it could be executed with only two people by using the wheeled case for the scan head as a "rolling dolly".  

May 29, 2001. Field Data Entry.

  After collecting all the scans we needed from our field trip we collaborated a plan for computerizing our field data, registering the scans together, and modeling the point cloud we had obtained. The process of computerizing the field data consisted of two initial steps: creating an AutoCAD drawing depicting scanner locations for the Amine Glycol Solution and the Carbon Dioxide Building and creating an Excel database containing the data we collected. The AutoCAD drawings consisted of a site plan showing each building, important piping, tank, scan, target tie point, and target sphere locations. The drawing also depicts the ScanWorld number, scan number and scan direction  recorded on site. These final AutoCAD drawings of the Amine Glycol and Carbon Dioxide buildings were then exported into the excel database file to be used in correlation with the scanning data. The excel database contained all the data we recorded during the scanning processes. We divided this information within the database into two parts, the Amine Glycol Solution and the Carbon Dioxide Building. Each part contained pertinent information about the scans taken, including the ScanWorld number, scan number, operator, scene range, sample spacing, number of points and the Kodak picture number. Any other important or unique information was also noted. Before we started the next step of registering the scans together, we found that the Cyclone Software allowed us to start modeling cylinders within one ScanWorld before or even during the registration process. This was achieved by opening the first ScanWorld alone to model the easiest sets of points that could be distinguished within the allotted point cloud, which in our case were the large separator tanks. As soon as we had the registration complete we could then proceed to turn off the point cloud visibility within the first scan and copy all its previously modeled objects into our final Registered Scan World, which would then contain a complete 3D model point cloud. This process of coping models from one ScanWorld to another proved to be extremely easy and efficient.

May 30, 2001. Annotating Registration Points. 

  Annotation is one of the most important steps in the process of going from the point cloud in the scans to ultimately creating a complete three-dimensional computer model of the Amine Glycol Solution area. First of all we added annotations to the registration targets and spheres that were collected from the scans in each ScanWorld. The annotation labels a specific point within a ScanWorld, so it was inherently important to have that same point in a different ScanWorld labeled properly with the same annotation label (or name). These Registration Labels tied the ScanWorlds together but again they must have been assigned correctly or the registration would be invalid, making good field notes essential at this point.  In the Amine Glycol Solution area alone we took ten ScanWorlds with registration targets in various locations throughout the scenes.  In each ScanWorld we placed at least three mandatory targets and often six to eight targets to increase the likelihood of an accurate registration.  If registration labels tying two or more ScanWorlds were mislabeled or switched an unacceptable amount of error would occur when attempting the registration.

  With all the Registration Points annotated, we could now create a registration. We first placed all the ScanWorlds into the registration and tried registering them all together at the same time. This was not a success on the first try and the error received was too great. One possible reason for this occurrence was a mislabeled target.  At this point we searched each ScanWorld looking for the mislabeled target. We might have simply switched two labels or had forgotten a label during annotation. We tried registering the ScanWorlds again yet there was still too much error. A different method of registration was needed.  From this point we started out by registering only two ScanWorlds and then we checked the error. This method did work so we tried adding ScanWorld Three and then ScanWorld Four and so on until we found the specific world that was producing the error. We find out that there was not enough targets to tie this ScanWorld Nine to the others, the reason for the reoccurring error we were receiving. Through this we find out that good planning when setting up the targets and spheres in the field was critical. So we then proceeded to find other objects in the scenes to act as tie points. We decided the top of one of the separation towers would work perfectly. This involved modeling the top of the same tower in each world and annotating it.  With this complete, we registered the ScanWorlds again and checked the error-which was now minimal and thus acceptable. We now came to the final part of registering- so we created a ScanWorld of the entire registered scene.  Everything had gone well until this point, however we must now create a model space in the ScanWorld.  For some reason this produced an Error Message. After much investigation we took ScanWorlds 9 and 10 out of the registration and registered them separately, the newly created model space worked fine. 

May 31- June 1, 4, 5, 2001. Modeling.

  After the registration, it was possible to get almost any viewpoint desired, therefore manipulation of the point cloud allowed for maximum visibility of the actual forms of the Amine Glycol Solution. The process of modeling began with the largest objects, being the tanks and large piping, to the smallest desired piping. The actual modeling of the objects consisted of obtaining the perspective of the form desired with the least amount of "stray points" that were located either in front of or behind the object. This at times became very troublesome but was resolved by rotating the perspective many times to receive the best vantage point. After this perspective was acquired a fence was placed around the desired form, the cloud was segmented by the fence, and a new window was launched where the selection was copied into the new model space. Within this new model space we edited the form without changing the original model space, therefore the models could be copied directly back into the original model space. After all "stray points" were deleted within the new model space, the point cloud was selected (picked) and then an object was created on these points by selecting from the menu toolbar: create object>fit to cloud>cylinder. The maximum absolute error of the object was checked next to see if it was less than .001. If the error was too great the modeling process could be repeated in order to try to obtain more points of the actual object and less "stray points". We repeated the process of fencing, launching a new model space, fitting a cylinder to a cloud, and copying the new object back into the original model space for every desired form and object. If there were two (or more) identical objects, we modeled the first object, copied it, and moved the new model into the point cloud of the second object. If certain pipes were connected by an elbow connector, we found that it was a lot easier and much faster just to let the Cyclone Software automatically create the elbow rather than trying to model it ourselves by using the point cloud. This was done by selecting both pipes and then selecting from the menu toolbar: create object>piping connectors>elbow connectors. Since this elbow had predetermined properties, we found it necessary to change the object properties to allow the elbow to fit within the correct point cloud. This was achieved by choosing from the menu toolbar: edit object>edit properties. Within this window the Bend Radius/ Pipe OD was changed from the preset 1.5 to .9, which we found to work for almost all our elbow connectors, though the value could be changed to any within the given range to correctly represent the point cloud being modeled as an elbow.

Modeling Patches: 

  Our objective at this point was to modeling patches from the point cloud of the building shell that is the generator shed. We first launched a new model space of the selection we wanted to work with and then deleted all unnecessary or "stray" points. The most successful way we found to achieve accurate patches that were representational of walls, roof, and floors was to make two small patches and then merge those selections together. It is extremely important, we found, to be aware of the points obtained that were to be converted into a patch, if "stray" points exist then the patch would average that data with the corrects points and the patch, when compared to the reality of the building, would be erroneous. After we created a correct and accurate patch, we made the it rectangular or circular with the specified number of sides and dragged the corners to the edge of the point clouds. We then copied and pasted the patch back into the original model space.  After repeating this process several times, we extended the patches so that they all met to form the shell of the generator building. 

Modeling Steel Sections:

  Modeling a small steel section became difficult in situations when the point cloud was only one inch apart and there were no other clouds registered of that same object.  Seeking out an unobstructed view with plenty of data was important. We learned as we went where to cut the clouds out and how to find the best area for the greatest results.  After modeling the steel sections, we copied them at an offset and moved them into place, within the point cloud, for the six identical structures. 

June 6, 2001.  Scanning Carbon Dioxide Building.

  The Carbon Dioxide Building went similar to the Amine Glycol Solution area with the only difference being that we had to tie the inside to the outside of the building.  We spent time making for certain that targets were visible from each scanner position both inside and out.  We had a great deal of piping crossing every which way so placing targets and spheres was a bit more difficult than before, but all in all the scanner process went extremely well and quickly.  The auto-acquire did not work as well in such a visually noisy environment so we "hand picked" the targets and spheres, which proved to be the most difficult and time consuming thing with which we had a problem.  In seven hours we took eleven scans including set up time.

June 7, 2001. Annotating and Registering.  

  The annotating process was the same as it was for the Amine Glycol Solution.  We later realized, back in the office, that could have taken a little better field notes, but the registering went well anyway. We were able to register nine of the twelve Scan Worlds together fairly quickly and accurately. Previously when trying to hand pick some of the targets and spheres, we had missed a few that later had to be found in the completed Scan Worlds, modeled and annotated. 

June 8 -19, 2001. Modeling.

  The modeling we did at this point was of the smallest pipes that we had previously decided to record. During the trip back to the Helium Plant site to record the Carbon Dioxide Building, we took pictures of the areas within the Amine Glycol that we didn't already have information for within the point cloud. This was done in order to be able to distinguish the locations of certain objects we knew we needed. With the pictures of the pipes, pipe connectors, tanks and other objects, we completed the unfinished elements of the 3D model. At this point we could begin modeling objects for the Carbon Dioxide Building. This process of modeling was the same as described previously.

June 20, 2001.  Scanning Separation Building.

  Scanning had to start early in the morning at the Separation Building because it is about three-hundred feet long and where the majority of the helium extraction process took place.  Planning was essential.  We used our planning sketch to set up the targets and repositioned them accordingly.  We (Glenn Hill, Jared Wright, and Jon Gamel) took a total of twenty scans.  Because of the proximity of other equipment, scans had to be planed well in order to capture the entire building inside and out.  We took thirteen exterior scans and tied them to the seven interior scans we took.  The total number of targets we had was our major problem.  We worked around this by placing flat targets on the walls of the building in triangulated patterns that were visible from multiple ScanWorlds.  This was our third time to use the Cyra Laser Scanner at the Exell Helium Plant so we worked well together and scanning went quickly.  We worked a total of twelve hours so scans averaged about 35 minutes a piece.  This included  taking the scanner to the roof top of an adjacent building for two scans.  Acquiring targets after the initial scan gave us trouble.  With in the Separation Building there are thousands of pipes and elbows going every direction and the scanner would continuously pick up piping and even light fixtures that it thought were the spherical targets.  We found that by using the highly reflective target on the back side of the spheres that time taken to acquire targets was cut in half.  We also found that targets more than 200 feet away were difficult to pick while acquiring.  This simply reinforces the idea of over targeting an area to avoid problems with registration.

  We simply choose the vertex of the target and give it a unique name.  This is another place where planning and documentation make a great difference.  The field sketches must be used so that the same target captured in different ScanWorlds can be appropriately named.

  Registration is simple if every ScanWorld has three common targets with at least one other ScanWorld.  It is also good if common targets appear in multiple ScanWorlds.  Here  we had the common problem of missing one target to register two ScanWorlds together.  The solution to this problem is to add a vertex to each ScanWorld that is missing target for registration.  The position of this new vertex should be in a place that is defiant like the corner of a door.  Some regular object can also be modeled and used as a target.  

  Modeling the separation building went similar to the Carbon Dioxide Building, however most of the equipment that the separation building once housed had already been removed or destroyed.  Our largest task was modeling the "skin" of the building and the concrete pads on which the equipment once sat.  The Control Room on the South end of the building where the entire helium extraction process  was monitored remained mostly in tact.