OpenStreetMap
I have been seeing new Bing imagery show up in Michigan in areas I edit. Although updated imagery is welcome, I would say my subjective view of its quality makes it a mixed situation. Subjectively I would describe it as more blurry or less distinct. Maybe the new imagery has lower contrast.
The problem I see most is with creating new or recreating existing imagery offsets in JOSM. I have created and uploaded many by using survey marks or benchmarks with known coordinates. An example that is no longer visible is at N 42 55 06.65005 W 084 32 34.97447. Contrast that with the Mapbox or Ersi World imagery where the concrete circular base is visible along with witness post just to the west.
I have started to add offsets for Ersi World which seems to have more distinct imagery. For the few offsets that I had added I have been quite pleased to see how close Ersi World was to begin with. Most have been within a meter accuracy.
讨论
GinaroZ 于 2017年10月28日 19:39 的评论
Yeah the ESRI imagery isn’t as good quality, but in most places it is newer or the same age as the DG imagery. Useful for updating areas marked as construction. With offsets it’s normally a case of aligning back to Bing, especially if there’s no GPS traces to help.
freebeer 于 2017年11月 1日 09:24 的评论
Hi, Without knowing where in Michigan you are mapping, here are some of my observations: The Bing imagery can consist of zoomlevel 20 imagery like around metro Detroit, easily identified by distortions in trees and some structures by attempting to re-align angled perspectives to appear to be straight on from directly above, something that can take time to gain experience with, to interpret correctly. Earlier Bing imagery at zoomlevel 19 was rather older, and now has been mostly replaced by lousy but newer imagery as you have noted. The older imagery was from aeroplanes and in some parts of the US had a noticeable offset from USGS data or state data, but at least was corrected, if not like the z20 imagery which seems to attempt to use LIDAR or something (I don’t know what that is so here I’m writing without knowing of what I speak) to correct things not at ground level (buildings, chimneys, powerlines and so), but at least based on ground contours. I don’t think that paragraph makes any sense, and isn’t what I intended to explain ;-) The newer Bing imagery appears to be of satellite origin, with native resolution far worse than the aeroplane images. At best zoomlevel 18 or maybe more likely 17. It’s artificially overzoomed in a way that does not show pixellation like with DigitalGlobe to be able to be viewed at ``native’’ z19. I’m not the only armchair mapper to have commented on the poor quality. In many areas I’ve covered, the previous old-yet-clear z19 Bing imagery can still be seen via ESRI at some zoomlevels, with a lower zoomlevel giving newer imagery. Despite being out of date I prefer to trace since-demolished buildings from that into empty areas, only to add a note once I discover newer imagery shows something different in its place. The Bing imagery, like the two DigitalGlobe sources, when from satellite, may be nearly aligned to other imageries, or (as nearer to me, commented by other mappers) may be many metres askew. But being so miserable, and meself having become spoilt with ESRI zoomlevel 21 source imagery in several parts of the US, I can’t bring myself to trace from anything Bing apart from z20, itself somewhat long in the tooth now. Perhaps it attempts to correct for terrain somewhat, but not enough for my standards today. In one particular area of Michigan the ESRI z19 imagery shows winter leaf-off conditions, while Bing z20 shows full leaf cover where there might be buildings. However if there is a terrain model to correct, in hilly areas there can be a wild difference to Bing z20 over a single city block, so there I trace missing buildings and individually align them to Bing z20 (that is, when I could have been arsed early on, when parts of the buildings could be made out under the trees). After seeing the Bing z19 offset nearer to the US east coast, I found there the best imagery to be then the USGS Large Scale imagery, before that service was castrated to its shadow of its former self today. Still, that remains my reference for alignment, me assuming the USGS will be suitably interested in correctness regardless of how blurry. In general I find that to be true enough, within the metre or so of blur present today, in all parts of the US I’ve been active, even where z21 imagery is available. My mapping causes me to cycle regularly through all available imageries, to guess at relative ages and offsets, to discover the different imageries can have varying offsets even over the expanse of a warehouse, so that I almost ignore the offsets so long as existing features are relatively close (highway centrelines, building outlines), given the errors induced by perspective and tracing inaccuracy likely are just as much. Man, do I write a load of old rubbish in far too many words. And that’s with fewer free beers so far to make the extended holidays yesterday and today tolerable. So far.
freebeer 于 2017年11月 1日 09:28 的评论
Sorry, first diary comment ever, there used to be paragraph breaks like what get preserved in changeset discussions, but no longer present above. Another hint I should keep things short and sweet. Rather than learn how to markup.