Funny how things shake down isnt it?! You choose a new car which has a certain rarity value and then see three in short order! You take ages to think of your next LiDAR News article and then three strong stories surface all at once. It is sometimes referred to as The Rule of Threes which generally applies when you wait ages for a bus and three come along all at once, but less commonly, when it relates to ones own thoughts, it is known as Confirmation Bias. Confirmation bias refers to a type of selective thinking whereby one tends to notice and to look for other instances of things that are already on ones mind.
So just as I stumble across an engaging topic for an article that spotlights measured building survey accuracy I notice that related Scan-toBIM articles are everywhere I look this month. Geoff Zeiss posted his article on the subject and referenced another he wrote a year earlier contemplating the Challenges in deriving engineering information from "scan to BIM" models. AECMagazine.com has in the past 3 months published not one, not two, but three features on the subject. In February 15 The Numbers Dont Lie addressed the savings to be made by accurately scanning an existing environment before using reality computing to prototype and track the design as it unfolds on site. In January 15 AECMag featured Scan-to-BIM and queried the need to laser scan every building, citing time and cost as a barrier in many cases when compared to traditional survey methods. And then in March 15 AECMag published a piece titled From Point Cloud to BIM in which the need for, and ability to model, absolute accuracy was called into question; favouring on occasion a more generalised accuracy as the appropriate alternative for some situations.
Not only does the cluster of above articles confirm my bias, it also coincides with some tests the GeoSLAM team has been doing in Nottingham, England. GeoSLAM is the developer of game-changing survey solutions for the measurement and mapping of multi-level three-dimensional environments. In partnership with with CSIRO (based in Australia) GeoSLAM has pioneered the design and manufacture of ZEB1 (a lightweight hand-held laser-scanner which records more than 40,000 measurement points/second as you walk); has commercialised the GeoSLAM Cloud (mind-blowing SLAM Simultaneous Localisation & Mapping software which automatically registers the survey data online); and has delivered a low-cost pay-as-you-go processing and download service to eliminate all upfront software costs and annual maintenance charges. I digress, but its important to explain from where this new technology comes.
The GeoSLAM tests set out to compare the accuracy of a measured building survey using traditional tape & Disto equipment with the accuracy of data recorded by ZEB1. To avoid even subconscious knowledge transfer three people were deployed: one person completed the tape and Disto survey and drew up the building plan in AutoCAD. A second person walked around the building with ZEB1, measuring and mapping every space before uploading the data to the GeoSLAM Cloud for processing and registration. A third person then downloaded the 3D point cloud and traced over the points using MicroStation to document the building plan.
With such diverse methods for measurement and drafting one may reasonably expect non-identical results. For example, a tape measure may sag slightly across its length, adding a few mm here and there. A Disto may not be resting on a perfectly flat surface and may be firing its laser beam at a slight angle from one wall to another; again adding a few mm. Such human errors (we cant blame the tape for sagging if it is allowed to sag by the person(s) holding it) are expected and they are accounted for. Many measured building surveyors will round up their tape measurements to the nearest 5mm or 10mm; perhaps less so if using a Disto which accurately and precisely displays the measured distance. Thats why the Royal Institute of Chartered Surveyors (RICS) in London publishes its own guidance for surveyors and survey clients so that expectations can be communicated and agreed before contracts are let. Included in the RICS guidance is the Survey Detail Accuracy Band Table Ill write more on that in a short while as I analyse the results in real-world context.
Example Measured Building Survey Results
The example building for the survey was an industrial unit spread over two floors with some double height space. Total floor area was 370m2 or 3,700ft2 in old money. It took the chap with the ZEB1 just under 15 minutes to slowly walk around the building, going into every one of the 14 rooms or spaces. 25 million measurement points were collected and processed by the GeoSLAM Cloud before being made ready for download at a cost of 10 or $15 (USD). In stark contrast the person with the tape and Disto spent a little over 2 hours in the building, measuring each room and the features within. Thats a site time difference of approximately 10x. Some vertical heights were taken with the tape or Disto but some features at higher level went unmeasured due to access and safety issues. The time taken to draw up each plan was similar and the resulting CAD layouts were visually compared. Immediately it became clear that the person with the tape and Disto had forgotten to recorded some features; entirely missing one column and miscounting the number of steps in a staircase. The ZEB1 had of course captured everything.
Randomly dimensions were also taken from each CAD layout in the native CAD applications and again the dimensions were compared for consistency and the differences recorded. In every case the difference from one CAD survey to the other was smaller than the width of the line when plotted at 1:50. Some of the differences could be attributed to user error where two people tracing over the point cloud may have derived different results just as two people measuring the space with a tape and DISTO may record slightly different measurements. Well never know; but we do know that:
- 12/15 dimensions or 80% were different by less than 10mm.
- The largest deviation was 0.13%; equivalent to +/-13mm over 10m.
- 2/15 dimensions or 13% were 100% accurate with 0mm deviation.
One way to check the deviations for relative accuracy is to add up all dimension lengths and compare the total length with the total of all dimension deviations. Using this method for analysis the total length was 115,515mm or 115.5m, and the total deviation was 116mm resulting in an average deviation of 0.10%.
Another way to check the deviations for relative accuracy is to add up the total deviation using + and – delta figures [i.e. (+4)+(+7)+(-6) = 5] to arrive at the SUM of all deviations. In this case the total length measured is still 115,515mm or 115.5m but the SUM of all deviations is just 68mm or 0.04%.
If we now apply these average deviation values to our average lengths measured (115.5m divided by 15 measurements = 7.71m) we find with Analysis A the average deviation is +/- 7mm and with Analysis B the average deviation is + 4mm.
RICS Survey Detail Accuracy Bands
Now to take the results and apply these to the RICS Survey Detail Accuracy Band Table wherein the accuracy bands range from A-to-J with provision to customise additional bands for any given use. As you can see in the image below, the accuracy band table sets out acceptable tolerances for Plan Accuracy and Height Accuracy. For each band the table also includes a brief description of Survey Types/Uses and it references the Legacy Plot Scale to which each accuracy band would be deemed appropriate. Finally the table includes a Minimum Feature Size that you would expect to capture.
In the RICS accuracy band table Plan Accuracy is divided into two columns, 1 Sigma, and 2 Sigma where 68% of normally distributed observation residuals will fall within the band value shown for 1 sigma, and 95% will fall within the 2 Sigma value.
In the GeoSLAM test case it was found that 80% of the dimensions were +/- 10mm and 93% of the dimensions were +/-20mm. The average dimension deviation (across all dimensions taken) was +/-7mm. Accordingly, these results position ZEB1 accuracy and usability between accuracy bands C and G in the RICS table. To add a little meat to those bones, in Accuracy Band C the ZEB1 is well suited for Heritage Recording of Buildings to create Plans and Sections. With intended usage at a Legacy Plot Scale of 1:20 ZEB1 is approaching the edge of its comfort zone. However, it is not so well suited for capturing the minimum feature size of 10mm.
In Accuracy Band D the ZEB1 is very well suited for Measured Building Surveys & Determined Boundaries. With intended usage at a Legacy Plot Scale of 1:50 ZEB1 is well inside its comfort zone and inaccuracies will be barely distinguishable. The minimum feature size of 20mm is at the very low end of what ZEB1 comfortably captures. For best results when scanning boundaries; the surveyor should ideally walk within 10-15m of the boundaries being scanned. To capture the minimum feature size the surveyor should walk within 5m of the features. The survey should ideally last no longer than 10-20 minutes.
In Accuracy Band E the ZEB1 is well suited for Measured Building Surveys, Low Accuracy Setting Out, Net Area Surveys & Valuation Surveys. With intended usage at a Legacy Plot Scale of 1:100 ZEB1 inaccuracies will be indistinguishable when plotted. ZEB1 will comfortably capture elements with minimum feature size of 50mm.
In Accuracy Bands F & G the ZEB1 is well suited for Low Accuracy Measured Building Surveys. ZEB1 will comfortably capture elements with minimum feature size of 100-200mm.
While accuracy is subjective, it remains imperative. Every survey needs to be appropriately accurate for its intended use. Defining the appropriateness of the accuracy will always be a matter for the client and his or her team. Selecting the right measurement technology and process will enable every team to balance their needs for accuracy with the imperative for speed. With time on site reduced by a factor of 10x when compared to traditional survey techniques, this game-changing technology from GeoSLAM is quickly becoming a profit-enabler for surveyors competing for business. For surveying firms enjoying 10x less time on site, GeoSLAM solutions may help to accelerate business growth. For other firms still trying to compete with traditional equipment, GeoSLAM solutions may accelerate their demise