Statements about iGuide accuracy

accuracy
uncertainty
error
measurement
#1

Question: When I was Googling a few days ago I saw an article reviewing that iGuide system that referred to 99.6% accuracy, but did not cite where that figure came from. Is that an accurate number?

The 99.6% number can be obtained when you use the (old spec) 4cm uncertainty on the max laser measurement range (10m). But at 2m distance (small room) that would result in 98%.

After we updated the published camera spec to 1cm uncertainty, the above numbers would change to 99.9% and 99.5%, respectively. However, all of that is marketing-speak and does not have firm rooting in metrology definitions.

It is better to keep it simple and say that the camera measures distance to walls with 1cm uncertainty out to 10 meters distance.

Other factors affecting the floor plan accuracy include how well rooms are aligned and how accurately the drafting team places walls and those factors are harder to quantify, just like when using tape measure and graphing paper.

In metrology, measurement uncertainty or accuracy is specified either in absolute units or as relative uncertainty, e.g. 1cm at up to 10m distance or as 0.1% at 10m distance or 0.5% at 2m distance or as a value for a certain confidence level.

The term “confidence” in metrology has a very specific meaning. For example, a measurement system uncertainty can be characterized as having 1mm sigma. It means that if errors obey Gaussian distribution, then in 66% of cases measurement error will be less than 1mm. Without Gaussian distribution sigma is obtained from standard deviation computed from a large sample size using a specific formula, but the 66% statement cannot be made.

A system uncertainty can also be characterized as having 1mm 95% confidence level. This means that in 95% of cases measurement error will be less than 1mm, irrespective of the shape of error distribution.

For more fun facts about metrology error statistics you can read this paper.
ijmqe120029.pdf (284.4 KB)

#2

Thanks for this, Alex!

#3

To expand further and cover errors in floor area (square footage).

Let’s say we established that for a particular house the measurement error in distance for outside wall to outside wall is 0.1% (of that distance). Due to math rules, that means that relative error in floor area will be about twice that, i.e. 0.2%. Similarly, for 1% error in linear dimensions (for example, Matterport’s marketing accuracy), the area error will be 2%. Simple examples: 1.01 x 1.01 = 1.0201 and 0.99 x 0.99 = 0.9801.

#4

Thanks for expanding on that. It gives me context for when I talk to realtors.