5 Replies Latest reply on Jul 7, 2014 9:56 PM by B Hall

# Accuracy comparison of the length in USA between Oracle Spatial and ESRI UTM/SPCS

Not sure where the post went - but here is my reply:

First off, assuming State Plane as the most accurate model is "plane" wrong (pun intended). It is a planer projection, meaning that you are taking a "sphere" and trying to represent a part of it on a flat map. So essentially it is always wrong. UTM - same issue, but even worse as it is over a larger area, so more distortion.

You said you are storing decimal degrees, so I'm going to assume a lat/lon coordinate system, probably SRID (EPSG) 4326. If this is the case, then there is no projection - Oracle is using geodetic (center of the earth) math. So:

1. None. All geodetic calculations.

2. Again, same geodetic math with no projection.

3. Yes. Much better than anything based on a plane!

4. Are you looking for micron accuracy or...? What do you call - accurate? The spheroid used for calculations is just a math model, and of course flawed. That said, it is generally accurate enough for most purposes.

5. Figure 2? Where are these figures?

Yes, geodetic calculations - being much more involved - are slower. If you are working over small areas (say 1000 feet), planar distance calculations are usually sufficient. But go very far (such as the area of a state), and you become quite inaccurate. You have to weigh what is more important. Of course, if these things are rather static in nature - you can just pre-calculate the area / circumference / etc and store and re-use that.

Bryan

• ###### 1. Re: Accuracy comparison of the length in USA between Oracle Spatial and ESRI UTM/SPCS

Moderator, where is my modified poster?

I included two figures (two links) in the modified poster.

If there are anything that are not allowed in my poster, please let me know.

Thanks,

• ###### 2. Re: Accuracy comparison of the length in USA between Oracle Spatial and ESRI UTM/SPCS

village75 wrote:

Moderator, where is my modified poster?

I included two figures (two links) in the modified poster.

If there are anything that are not allowed in my poster, please let me know.

Thanks,

This is your very participation as a member/user of the OTN forum site.

Your post that needed to be examined as a legitimate post, and not something such as might be contrary to the Terms Of Use or the guidance given in the site's FAQ..

Every time you modified it, you threw the thread into an examination queue -- all over again -- same as when it first existed when you started it.

Do NOT modify posts that you have made.

Ever.

When you do that you destroy the original information and people that may be ready to respond will be answering to "nonsense".

ALWAYS add a reply to yourself with whatever information you hope to change and the reason(s) for such change, even if it is to correct a simply typographical error..

• ###### 3. Re: Accuracy comparison of the length in USA between Oracle Spatial and ESRI UTM/SPCS

I see.

Here is my whole story. I posted my first unfinished one in accident.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

We have many polyline features of the USA in the format of Shape file.

The Shape files are uploaded to Oracle database and the length of the features are displayed to the Web when required.

The lengths of the features are in decimal degree (Geographic Coordinate System) in default in Shape files,

However, Kilometers (KM) or Miles are required when to be displayed on the Web.

We think that there are two quick ways to convert the decimal degree to miles (if there are other quick ways, please let us know, we appreciate it):

1) Using ESRI ArcMap to re-project the Geographic Coordinate System (GCS) to one of Planar Projection Systems, e.g. UTM, SPCS.

Then, use ArcMap's "Calculate Geometry" function to convert decimal degree to meters in Shape file.

Finally, upload Shape files to Oracle database where the lengths are stored in meters.

The meters are directly used to display the lengths of the features on the Web.

2) Directly uploading Shape files to Oracle database where the lengths are stored in decimal degree.

When the lengths are required to be displayed, using Oracle Spatial (SDO_GEOM.SDO_LENGTH) to calculate the lengths in meters on the fly.

There is another way to calculate the length of any two points in the Earth surface directly based on Haversine Formula.

For deciding which the above ways is the best considering the accuracy and speed,

we firstly compared the accuracy of the Oracle Spatial by comparing with ESRI UTM.

We selected 461 features in Alaska (AK), 1105 features in California (CA), 1227 features in Florida (FL).
Figure 1 shows the accuracy of the lengths compared between Oracle Spatial and ESRI UTM.

We think that the State Plane Coordinate System (SPCS) is the most accurate coordinate system that applied to the whole USA.

Therefore, we take SPCS as a reference. It means that we use the lengths calculated based on ESRI SPCS as a baseline,

we compared the lengths calculated based on ESRI UTM and Oracle Spatial respectively.

Figure 2 shows the comparison results. It looks like that Oracle Spatial provides more accurate lengths than ESRI UTM while taking ESRI SPCS as a reference.

We did not consider the speed yet.

Now, we have below questions:

1)    What projection is used in Oracle Spatial, especially for SDO_GEOM.SDO_LENGTH?

2)    Does SDO_GEOM.SDO_LENGTH use the same projection as SDO_GEOM.SDO_DISTANCE?

3)    Anyone has an experience to use Haversine Formula? What about the accuracy comparing with current available other ways, e.g. Oracle Spatial, UTM, SPCS?

4)    Is there any other professional or open source software providing with higher accuracy?

5)    From Figure 2, what other conclusions we can get, especially the line of Oracle Spatial to UTM compared with other two lines?

6) What is the difference between Haversine Formula and Geodetic Calculation (after I saw Brian's response)?

• ###### 5. Re: Accuracy comparison of the length in USA between Oracle Spatial and ESRI UTM/SPCS

Bryan