Product Review by Meridian Engineering: FARO Scan Localizer

Product Review: FARO Scan Localizer

Thank you to Meridian Engineering for this article.
By Brian Boehmer and Tiana Moe

Last month Meridian had the opportunity to test out one of FARO’s latest products, the Scan Localizer, through FARO’s Early Adopter program. The opportunity to give feedback has greatly enhanced the Localizer since it first appeared. The product tested is not the same as it was last year. It has been improved by the feedback of the user community. While there are a few wrinkles that need to be ironed out, the Localizer is an undeniably useful tool to help streamline the scanning process.

The Localizer utilizes SLAM algorithms along with IMUs and other sensors to determine its position as it is moved around the environment. It is primarily designed to be paired with one of the X-series scanners, however it is now possible to use it in a standalone fashion.

Meridian’s LiDAR team spent some time with the Localizer, to see if it achieved at the level FARO has advertised. Meridian also pushed past what FARO recommended to test the limits of what it can do. Understanding the limits of any tool is what allows for the best use of it.

Process
The Meridian team wanted to see how much the Localizer improved registration results. The measures of the testing were the amount of time that could potentially be saved and the quality of the initial registration. To do this, they scanned 8 areas using the Localizer. The scans were then processed in Scene 6.2 using the preliminary transformations provided by the localizer. After resetting all the scan transformation values back to zero-zero (eliminating the localizer data), they reprocessed that same data set. All the processes were run on the same desktop computer with no other processes running. Testing only involved a single automatic registration, including a top down registration followed by a cloud to cloud registration, using default parameters.

Test Results

Test 1: Inside Office
Number of Scans: 9

_______________________With Localizer        Without Localizer
Preprocessing Time        15 minutes            12 minutes
Registration Time           10 minutes            25 minutes
Registration Results        Successful             Successful

The office scan consists of office space, cubicles, and hallways on a single basement level. It was scanned as an initial test of the basic functions and output of the Localizer. Before registration the scan placement was within approximately 1 foot of final placement after registration. The registration was tighter than it was without the Localizer. Testing was also done with the Localizer as a standalone tool to create a 2D floor plan. The exported map quality is of poor resolution, making it most suitable as a small thumbnail image. Having the option of storing the map at a better resolution could make this more useful as a standalone scan product. Regardless of resolution it is extremely useful as part of the user interface.

Test 2: Outside – Buildings
Number of Scans: 14

_______________________With Localizer        Without Localizer
Preprocessing Time        16 minutes            11 minutes
Registration Time           11 minutes            67 minutes
Registration Results        Successful             Failed

Test two was pushing the localizer outside, as well as incorporating some limited vertical movement. The scans began inside the office down an alley and around the front where there is a flight of stairs up to the parking lot. The 8-foot shift didn’t seem to bother the Localizer. The scanning continued across the parking lot to an office complex to the south and scanned in between the buildings.  The Localizer worked very well outside in areas where the combination of buildings, trees, light poles, and other vertical features were present. In such areas, it resulted in better initial positions than using the position provided by the GPS in the scanner.

Figure 2: Initial scan placement with localizer

Test 3: Parking Lot
Number of Scans: 5

_______________________With Localizer       Without Localizer
Preprocessing Time        5 minutes            12 minutes
Registration Time           6 minutes            28 minutes
Registration Results        Successful           Failed

Testing to see how few vertical objects were necessary for the Localizer to get its bearings, this test was performed in a small parking lot. The parking lot was quite full, and the Localizer worked using the cars to keep its location. It is interesting to note that the Localizer does not use the GPS data that is collected by the scanner. This successful test spurred an additional test to find the breaking point of the localizer solution. A wide-open parking lot, with no cars and very few landscaped planters was selected and the localizer unsurprisingly failed. The GPS in this case was more accurate than the localizer positions.

Test 4: Hallways and Multi-Level Scanning
Number of Scans: 12
_______________________With Localizer                  Without Localizer
Preprocessing Time        45 minutes                      48 minutes
Registration Time           30 minutes                      21 minutes
Registration Results        Successful (2 attempts)    Failed

This building has two floors with a staircase on each end of a long hallway. Each floor has multiple rooms off one side of the hallway. It was also decided to see if the localizer could maintain a solution if it was moved down a staircase. On several instances the Localizer lost connection with the scanner, didn’t know it was finished scanning the area, and failed to continue to localize when moved. This forced the creation of new localizer projects. This occurred a few times while scanning this building with no apparent cause. A glitch occurred just before the staircase but seemed to correct itself so scanning continued. The scans that were done in the stairwell did not adopt the proper transformations from the Localizer and were placed 4,000 feet apart in elevation, likely the altimeter elevation. This made the automatic registration impossible. Once those scans were removed from the solution, the registration worked well, but only for the localizer positioned scans.

Test 5: Hallways and Multi-Level Scanning
Number of Scans: 23

_______________________With Localizer        Without Localizer
Preprocessing Time        82 minutes            81 minutes
Registration Time           21 minutes            31 minutes
Registration Results        Successful            Failed

This was a continuation of the building scans of test four. The Localizer performed well
even with the large vertical shift on the opposite stairwell. The stairwell scans covered
four 90-degree turns and a vertical shift of 15 feet. The lower hallway had rooms directly
underneath the rooms above, although of different sizes. The Localizer positioned all
scans accurately in horizontal position however it grouped all the room and hallway scans
onto the same vertical plane. The scans in the stairwell did not use the altimeter data but
were positioned each slightly higher than the next by the localizer. This indicated that the
Localizer was detecting some sort of vertical movement. Scene had no trouble with the
scans being on the same vertical plane and placed them accurately after the automatic
registration.

Figure 3: Initial scan placement with localizer

Figure 4: Top down after registration with localizer

Figure 5: Front view after registration with localizer

Test 6: Maze-Like Room Configuration with Some Multi-Level Elements
Number of Scans: 25

_______________________With Localizer        Without Localizer
Preprocessing Time        60 minutes            61 minutes
Registration Time           40 minutes            62 minutes
Registration Results        Successful             Failed

This area consisted of thirteen rooms that connected one to the next without hallways. The room configuration spiraled in on itself in a rectangular pattern. The rooms were mostly on a single floor with a few rooms with a small number of stairs. The initial placement from the Localizer was very good, which lent to a quick registration. Without the Localizer, the scanner failed to come close to a correct solution.

Test 7: Auditorium with Balcony or Mezzanine Level
Number of Scans: 26

_______________________With Localizer        Without Localizer
Preprocessing Time      45 minutes            45 minutes
Registration Time         29 minutes            53 minutes
Registration Results      Successful             Successful

These scans covered a large conference room or auditorium with a mezzanine level, 37 feet in height. There was seating on the main level on two sides and around all four sides of the mezzanine. The upper level was linked to the lower with six spiral staircases. The Localizer did well with initial scan placement and the registration was very good. This area also registered well without the Localizer. Using the automatic registration including a top-down and then cloud to cloud processing. The scans had a high overlap percentage due to the open floor plan.

Test 8: Single-Level Scanning
Number of Scans: 15
_______________________With Localizer        Without Localizer
Preprocessing Time        62 minutes            65 minutes
Registration Time           11 minutes            14 minutes
Registration Results        Successful             Failed

Test eight included a hallway with two large rooms and several side rooms on a single floor.  The localizer worked smoothly, however the initial placement was slightly skewed. The registration went well despite the skewed placement.

Conclusions

In general the localizer worked well and performed as advertised. The only exception being the use of the 2D map as a standalone product. The usefulness of which is not apparent, at least with the current limits on registration.

•    Registration Speed
After the tests, the Meridian team found that, on average, using the Localizer cut down registration times by 49% compared to time spent registering without the Localizer. In one instance the Localizer processing time was 17% of the time it took to process scanning data without the Localizer.

•    Precision of Registration
When scans were registered using the localizer, the placement was more precise in terms of the mean point errors, compared to an automatic cloud-to-cloud registration on the same data set.

•    Order of Scans
Sometimes the individuals doing registration activities are not the same as those who are in the field. Scanning in a disconnected and non-consecutive order can confuse the non-field personnel. In the field, the scan crew often needed to scan areas out of order to avoid other people on site. Using the Localizer, the order of rooms wasn’t as important—they could skip a couple rooms and come back to them later, without providing confusing data to the office technicians when it came time to position the scans.

•    App Interface
The map interface worked well and was useful in the field. It was enjoyable to watch the slam algorithm continue to refine the solution. It was used a lot to deal with curious people who wanted to know what was going on and stop them from walking into the scans. Also, the interface indicated when a scan was finishing up, removing the need to peak around the corner and get scanned, or running a timer.

After testing the scanner the way it was designed the Meridian team decided to push the Localizer a bit further. And found a few pleasant surprises.

•    Outdoor Testing
The worked well even outside and didn’t need as many vertical features as was initially suspected.

•    Multilevel and Vertical Projects
That the Localizer would work well on multi-level projects was not an expected result. It was expected this would break the solution of the Localizer if it was taken up or down stairs. The initial transformation of the scans based on the Localizer eliminates the z values normally obtained from the altimeter. Being able to add the z from the altimeter in Scene or in the field could perhaps enhance the multi-level solution even further.

In trying out any new gear naturally not everything is going to be perfect on the first go-round. The Meridian team identified a few qualms that will hopefully be rectified in the next version.

•    Size and Weight
The Localizer doubles the weight of the tripod, making maneuvering from room to room more difficult. There is no need for a tripod dolly or to haul around a computer after the latest firmware updates (that would have killed interest pretty quick). The size is a bit more than double the size of the scanner. It had to be removed to scan crawl spaces or other similar areas.  It is annoying to go from something small and compact (the X-series scanners) to something that is big.

•    Glitches
The Localizer lost connection with the scanner on several occasions and there seemed not have any common reason why. Rebooting the Localizer helped on one occasion, but more often rebooting the Localizer meant it lost its position and map, forcing the creation of a new project. Clearly this was less than an ideal result, especially when the team was halfway through scanning a floor.

•    Quick Connect
The mount of the scanner to the Localizer is sturdy, but it is time-consuming, requiring the placement of four bolts. The current case configuration does not allow for the scanner and Localizer to stay together when stored in the case, which means it is being bolted and unbolted at least once a day. The repeated bolting and unbolting was a worry, it could lead to the bolt threads wearing down after repeated use. A quick connect and a change to the case would alleviate this problem.

•    Orientation
Currently it seems that the Localizer uses the initial direction in which the Localizer is facing as the top of the map. If the Localizer is not facing north, the top of the map is not north, which could create confusion. It would be a great plus if the localizer used the compass of the attached scanner for its initial orientation.