OpenStreetMap logo OpenStreetMap

Each imagery offset is different, and over time a patchwork of different referencing systems emerges… one big disaster, worse than “GCJ-02” (which is at least “smooth and not noticeable”.) Happens even with just one person editing. You’ll notice it when you finally join village A to village B. Editors of village A have been pulling the imagery downwards to match what they have already mapped. And editors of village B have been pulling the imagery upwards…

I hope one day, after offsets are finally no longer present in imagery, Artificial Intelligence (?) will somehow fix the mess.

Location: 基寮坑, Dongshi District, Taichung, 42351, Taiwan


Comment from kucai on 2 March 2020 at 07:17

Each imagery has different “zoom and skew” even if all the imageries are aligned to a surveyed point. In my experience for my country, Bing has the clearest/cleanest image, then Esri then Maxar. But Maxar Premium has the newest.

In this case, I align Bing to a ref point, then go to my area of interest. I find an obvious building to trace the outline, then I’ll align Maxar to this building. Trace what needs to be traced from Maxar and then return to Bing again.

Unless all imagery companies agreed to a single source provider, or agrees on a set of ground control points (GCP), there’ll always be discrepancies.

Comment from jidanni on 13 July 2021 at 17:33

Today I illegally tried my country’s government map as a background, and found all my years of local mapping was skewed… proven by GPS traces (which match the government map) too. Well that’s what happens after a constant diet of your average satellite images, and a government that doesn’t allow one to use the local imagery or maps as reference.

Comment from kucai on 14 July 2021 at 02:40

Hah, see! it’s a never ending cycle of offsets. On another note, there is an area south of my country where the images seemed to be offset. I happened to do a survey job in that area, so I managed to get cm accuracy data and also corrected drone imagery from that survey. In addition, JOSM just added the specific carto projection for that state. Surprise, surprise! the original old Bing imagery actually lined up within 1-2 meters with the drone imagery using that coordinate projection. Makes you wonder how many more aligning mistakes were made out there.

Log in to leave a comment