4 ms·
A lot of people are misunderstanding the article in this thread. I work with GIS data from multiple sources. Neither Google's APIs nor OSM nor Bing nor anythin
by plonk 4y ago
A lot of people are misunderstanding the article in this thread.
I work with GIS data from multiple sources. Neither Google's APIs nor OSM nor Bing nor anything else is enough for applications with high specs. You need to mix machine learning and available sources.
Doing this is a PITA. Sources rarely match one another (e.g. an object will be referenced on both Bing and OSM, but 20 meters apart, with no way of accurately matching them in the middle of a dense city).
The article very clearly says that the goals include making the existing services interoperable, which would be a big advance for API customers. (This isn't about making a new Google Maps competitor, it's about a Google Maps _API_ competitor, though it will surely help build the former too.)
Just being able to download the same area from Bing and OSM and seeing "this house has ID 234 here and 234 here, that's the same building" would be great, because that lets you gather all the available information about it instead of having to choose one or use matching algorithms that can make mistakes. It can also remove false positives, which also happen.
- secretsatan 4y agoI’ve been dealing with this recently and it’s a mojor headache, but the issue of things not lining up is primarily one of using coordinate systems correctly, knowing what things were measure in and what they are shown in
- plonk 4y ago> but the issue of things not lining up is primarily one of using coordinate systems correctly, knowing what things were measure in and what they are shown in Not in my experience. Two satellite/plane pictures from the same provider can be shifted by several meters. Buildings hand-drawn in OSM are rarely perfectly correct, and won't fit most images anyway due to the first point. Conversions between coordinate systems are precise up to floating-point errors, that's the easy part and more or less solved by geo-referenced data formats.
- secretsatan 4y agoYes, but i’m finding, increasingly, that while we have the means to accurately record data, even accounting for float point errors by reducing area of coverage, things are simply not reported or cared about. It took us a considerable amount of investigation in using an rtk system to actually set things up correctly to get it to actually work within it’s 1cm accuracy. This is stuff like contacting rtk service providers to find out what system they actually correct their coordinates to. To me, the greatest innovation any provider could do is just specifying the coordinate system used EDIT: in fact i am finding it shocking, it’s even invoking government regulation, the simple fact is, even using advanced systems from leading suppliers, vital facts are just not recorded because geographic coordinates system are so numerous and complex suppliers just offer defaults and do the real work in the background Even here, i hope you don’t misunderstand a geographic system as we talk about all coordination systems, as opposed to geographic vs projected coordinate systems
- plonk 4y agoWhat data sources do you work with? I didn't know about RTK, it looks interesting but seems hardly applicable to satellite imagery, since it seems that you'd need a lot of receiving stations on the ground? My understanding is that providers like Maxar report coordinate systems correctly, but that errors in ortho-rectification (due to imprecise or wrong terrain altitude data) make it hard to make two images of the same area match exactly. > even using advanced systems from leading suppliers, vital facts are just not recorded because geographic coordinates system are so numerous and complex suppliers just offer defaults and do the real work in the background Not sure I understand what you mean by this. Do you get data that reports a coordinate system different from the one it actually uses?