Author Topic: Reverse geocode and city  (Read 3605 times)

Offline HvTFoto

  • Newcomer
  • *
  • Posts: 30
    • View Profile
Reverse geocode and city
« on: May 30, 2019, 05:16:39 PM »
Hi,

I just upgraded to PM 6 and started to "play around" with PM 6 Plus (build 3215). I've created a catalog with some photos. Some of the photos have been shoot at my summer house in Norrtälje and contains GPS data. When reverse coding these photos the "City" field states the location as "Sandviken." (some 200 km NNW of Norrtälje). When going to Image Information and clicking "show on map" the location is correct (i.e. shows the place in Norrtälje where the photo was taken). Another photo shoot at my brothers place shows the street name as the city and the street number as the location so there is definitely something that isn't right with the database and/or the reverse geocoding. I attach one of the photos (a handwritten note containing information on colour used when repainting the house) to illustrate the problem.

Regards

Hans

pablogg

  • Guest
Re: Reverse geocode and city
« Reply #1 on: May 31, 2019, 08:56:49 AM »
Thanks for the feedback Hans. Looking at the raw data returned from our OpenStreetMap database, it is returning the following for your coordinates (see below).

Since OpenStreetMap does not always provide the "City" keyword, we have had to "guess" which field to use to denote City. Our current logic is to check the following fields (in priority order): [city, town, suburb]. In your case it is defaulting to using "suburb" for the city.  Looks like we may need to add "suburb to the city priority list [city, town, "county", suburb] so that it correctly selects "Norrtälje". Will review with team to make sure it makes sense globally to make this change.

Note: if anyone else has examples of incorrect reverse geo results, please post so that we can get as accurate as possible with our geocoding. These changes can be made on the server, so will not require a new PM release.  Thanks.

Regards,
-Pablo

<reversegeocode timestamp="Fri, 31 May 19 08:37:37 -0700" attribution="Data © OpenStreetMap contributors, ODbL 1.0. http://www.openstreetmap.org/copyright" querystring="format=xml&lat=59.74033&lon=18.65358&zoom=">
<result place_id="175156073" osm_type="way" osm_id="431568212" ref="Flyhagsvägen" lat="59.7397281" lon="18.6490506">
Flyhagsvägen, Sandviken, Campus Roslagen, Norrtälje, Sveriges Landskap, Stockholm County, Region Svealand, 76146, Sweden
</result>
<addressparts>
<road>Flyhagsvägen</road>
<suburb>Sandviken</suburb>
<village>Campus Roslagen</village>
<county>Norrtälje</county>
<state_district>Sveriges Landskap</state_district>
<state>Stockholm County</state>
<postcode>76146</postcode>
<country>Sweden</country>
<country_code>se</country_code>
</addressparts>
</reversegeocode>

Offline HvTFoto

  • Newcomer
  • *
  • Posts: 30
    • View Profile
Re: Reverse geocode and city
« Reply #2 on: May 31, 2019, 10:45:01 AM »
Thanks for the prompt reply. I do have a photo from a different location (westcoast of Sweden) with a similar problem. I can provide you with the photo if that would help you in correcting the problem. When it comes to the photo I sent earlier I have the following comments:

(1) <suburb>Sandviken. There is a camping site some 30 km from my location which is named “Sandviken camping” but it has nothing to do with “suburb”, so I don’t understand why that comes up in that field,
(2)<village>Campus Roslagen. According to my understanding (English isn’t my first language) village means “small town” and Campus Roslagen is not a “village” but a small area in the city of Norrtälje. Campus Roslagen is situated some 4 km from where the photo is taken.

Hopefully these comments will help in deciding what changes has to be done.

Regards

Hans

pablogg

  • Guest
Re: Reverse geocode and city
« Reply #3 on: May 31, 2019, 11:31:05 AM »
Thanks Hans.
Just so you know, the database used by the PM product to perform reverse geocoding is not created by Camera Bits. It is an open source (publicly available) database that we simply read.  So we are limited in what we can do to correct any potential  errors in the data. For this case, it seems to make sense to include "county" to match city. However, we need to make sure it satisfies most other situations.
Regards,
Pablo