Querying 2D-geometry data

[manuschiller]

I got a huge amount of tracking data from a research study (26 Objects * 25 frames per second * 90 minutes * 50 = ~175 500 000 records in my Database).

The reference system of the tracking data is a fixed environment of x*y meters, and not a geographical lat / long system.

The schema (MySql) looks like this:

TABLE trackingdata:
uid(int) 
object_id(varchar)
coords(POINT(X Y)
voronoi_cell(POLYGON(...)
voronoi_area(float)
frame(int)
timekey (datetime)

I want to use elasticsearch to efficiently query and analyze (love kibana!) my data.

Since I don't have much experience with elasticsearch i was wondering if it can handle geometric data (POINT, POLYGON) or if it is restricted to to its gepoint-datatypes (geographical, not geometric) datatypes I found browsing the documentation ( GeoPoint, GeoShape )

Would elasticsearch be capabable of finding records by an example query like this:
"find all voronoi_cells that are within a rectangle described by coords with a specific voronoi_area"

[spinscale (Alexander Reelsen)]

Hey,

IIRC the geo points/shapes are geographic for now. If it is possible to convert your data, you could use a custom tile server to visualize your data like in this example with the world being an ice hockey field https://www.elastic.co/blog/kibana-and-a-custom-tile-server-for-nhl-data

--Alex

[manuschiller]

thanks mate! that helped a lot

[manuschiller]

thanks again for your input. The visualization itself works like a charm with a custom tile server. But I am wondering if I can use all the query filters by elasticsearch for my cartesian 2D-data. Would I have to transform the data to the geodetic system (is that even possible?) or would the results be valid for my data as well?

[spinscale (Alexander Reelsen)]

Hey,

you would have to do some transformations yourself between your own mapping and the real world, so that the tile server approach works. This would have to be done on index and query time.

Hope this helps.

--Alex