Wednesday, July 6, 2016

Reliability of the calculation of total distances and accumulated heights: original versus optimized tracks

We reproduce here some questions from one user related to the reliability of the calculation of the height and, also, cumulated climb and descent in the case of both the original track and the optimized track. 

Also, we will seize this opportunity to give a brief overview of our error treatment processes.

A summary of the questions of our user:

How reliable is the calculation of the distance and height difference your tracks and traces optimized?

I would like to know the most reliable figure for the distance and the accumulated climb in the case of the attached track.

Also, when I open the link " info points list" on the right side of the table which appears there,  there is an "optimized trace" with a cumulated climb of 537.43m. 

As I understand, your system detects the GPS system's recording errors and also corrects the optimized trace?

So in summary, the result obtained is generated from the corrected trace.

In this case, I note that reliability of the original track is D (= bad) and correction level 3.

And the final result for the distance and cumulated climb: 14.143 km and D + 466 to 469m.

Is this the right answer?

Here is a summary of our responses:

Indeed, 14,143 km and D + 466 to 469m are the figures describing your track

On this screenshot, we can see that for this part of the track...



















was recorded with this profile by the GPS device.
























That is why our system has applied the corrective algorithms.

One can intuitively see that it is very unlikely that the track’s original data, without the corrections we will write about further in this post, is describing the real profile of the physical path traveled by hikers.

We take the opportunity of this example and your questions to describe our error handling process briefly.

This is a schematic description of the data processing process IBPIndex:

1. Resolution of different problems caused by the GPS recording device in the .gpx track "original" essentially:

• Analysis of abnormal points

• Jumps / Steps in the profile of heights product, for example, of stops on the route or geographical elements that blocked or prevent a proper reception of the GPS signal

• Optimization of the number of points recorded on the track: the more points are recorded, the bigger the accumulated mistake is introduced in the calculations.

• Other error treatments methods...

…and we obtain the optimized track

2. On the optimized track, we apply a series of other correction mechanisms. 

The main one is an automatic correction system working in 8 possible degree levels ranging: from level "0" no correction, no correction only optimization only, up to level "7" maximum correction.

Once the data has been again revised/corrected we assign a reliability degree to the track, according to the following scale:

"A" = Very Good
"B" = Good
"C" = Fair
"D" = Bad
"E" = Very Bad

And after this final set of analysis and corrections, the "Final Track" is obtained.

This "Final Track" is the base on which IBPindex calculates the final statistics of the tracks that you see on our website: positive and negative cumulative altitude difference, percentage rise, etc.

Tuesday, July 5, 2016

Differences in accumulated heights among GPS devices, IBPindex and data analysis apps

The impact on the reliability of the tracks' statistics of the different error treatment methods is a hot topic.

The following question of one our users shows the practical impact of those methods, and how IBPindex is implementing much more sophisticated techniques that are closer to the physical reality of the trail.

Our user question:

I would like to ask you about the noticeable differences in heights existing between the ones provided by the GPS device and the ones produced by your analysis. 

I’ve noticed that usually, the number accumulated climb is bigger in your analysis than one given by the GPS, and I have checked this with two different devices. 

For the attached tracks and the same route: the difference in the accumulated heights between the two devices is just 5 meters, 1266 in one case, 1271 in the other, versus 1718 and 136ibp and 1759 and 137ibp in your analysis. In the case of the length of the track, the numbers are nearly the same in your GPS and your system.

Could you explain to me why this difference in the accumulated climb, and not in the lengths or distances?

Our answer:

We have analyzed the files you sent us and difference in the accumulated totals stems from the error treatments methods used by the GPS devices.

The GPS related errors depend on basically: the position and number of satellites present at the horizon during the time span in which the route is recorded and the precision in the recording of the heights.
You can find more information in:

http://en.blog.ibpindex.com/2014/07/accumulated-climb-and-descent-influence.html

http://en.blog.ibpindex.com/2014/05/cumulated-height-why-same-file-produces.html

The most usual error treatment methods consist in the application of heights meshes (see the link mentioned above) or to discard any variation in heights smaller than 5 meters (as you can see in this screenshot featuring the default options of a GPS data analysis software)
























The choosing of this error treatment method causes the following situations:

Let’s imagine a route of a total of 50km with constant ups and downs of 4.9m all along the road. For the cyclist would be physically very demanding but, if the GPS system discards all the slopes of less than 5m of length, you will see that the accumulated climb or descent amount to 0 !!

IBPindex uses an array of error treatment techniques much more advanced and precise than the previously mentioned truncations and substitution of the original heights, such as:

•    Detection and removal of aberrant points
•    Detection and removal of saw tooth patterns
•    Optimization of the number points fed into the analysis, to avoid that the addition of a big number of little errors generates a significant cumulated deviation.

According to our experience, those and other techniques, that we have progressively incorporated into our algorithms along the analysis of more than 2.000.00 routes, make our results given by IBPindex are the ones that describe reality the closest.

There are various open debates in Internet forums about the treatment of heights, on this link you can find another example.

http://es.blog.ibpindex.com/2016/05/tracks-ruta-monegros-bike-marathon-2016.html (in Spanish)

Please do not hesitate to ask us about any further doubt you can have related to these or other tracks.