OpenStreetMap logo OpenStreetMap

kucai's Diary

Recent diary entries

Cellphone photo GPS tagging

Posted by kucai on 1 January 2024 in English.

I have the suspicion that my phone is tagging the GPS location for some photos rather wrongly (timed offset is my guess). Found this out when comparing the locations against those indicated on JOSM photo overlay.

This despite the phone getting both L1 and L5 signals, and running with no sim card which should reduce cpu load further since there’s not much background processes that need to be handled.

To say I’m pissed is rather an understatement. Moving on, I will try leaving the phone to only record gps track and then matching photos from my cameras to that. No electronic shutter so extra wear and tear on that mechanical one! Shoulda bought that oly em10 back then.

Newish L5 gps phone testing.

Posted by kucai on 4 January 2023 in English.

Recorded a track using opentracker app for a 100km journey through the backroads. Phone was placed as far back on the pillion seat to get maximum view of the sky. No kalmann filter or anything.

The resulting track was very smooth, extremely few jaggies. One particular section of note is at the newly opened elevated highway that is probably at least 10 stories high. Due to the slanted imagery, the traced highway and the gps track differs by about 15 meters. Personally, I won’t bother realigning (although the temptation is very irresistible) - besides somebody else would probably moved it back according to imagery. I learned long ago that trying to get people to check imagery alignment with the provided cm level survey points before editing is a hopeless endeavor.

Still, I am pretty excited with the raw GNSS measurement ability of the phone and intend to set out more survey points just for the hell of it. Waiting for kind, knowledgable people to provide an appimage for the gnssanalysis software since I couldn’t get past installing matlab whatever in its current form.

Location: Kampung Sungai Jernih, Kajang Municipal Council, Hulu Langat, Selangor, 43000, Malaysia

So I got this message from JOSM:

Error message(untranslated): Placeholder node not found for reference -104469 in way 741879858

Tried various ways to fix the error but it was getting late, so I saved the whole thing as a joz session to tinker with it later.

Next day I tried to load the saved session. No bueno! josm wouldn’t load the data layer because of the error.

Arrgghh.. cue the rage meme. That layer got a complete bus route that I started from beginning to the end, with the accompanying highways chopping, poi adds along the way and all kind of stuff..

Decisions decisions..

Posted by kucai on 18 January 2022 in English.

GPS tracks versus image aligned to a surveyed position (cm level RTK) 7 km away - displacement of 20+ meters North South axis.

Surveyed position might have image provider problem, that is, improper stitching and alignment of photo imageries during processing.

GPS tracks has the problems of inherently inaccurate receiver, placement of phone during recording, and possible Kalman filter in the gps app.

Mappers have the problem of not knowing that only JOSM has the imagery alignment database.

Weird day in JOSM

Posted by kucai on 18 September 2020 in English.

Tried to upload edits in newest JOSM but validation crashed the program with out of memory. Allocated 4GB, but JOSM said gib more clay, so added 2GB more. Still failed. All six cores of my (uber powerful FX) were pegged at 100%, eventually I had to kill java. Tried with the previous version with the same result. Finally, after a few restarts, I uploaded my edits without any validation checking. That’s a first for me.

It’s kinda strange. I only got about 800 objects to upload and it failed. I regularly uploaded more than that, almost double, successfully while also fixing all the validation flags that I can fix. The only thing different that I can think of is that I used the remote control feature for fixing a couple of notes from main openstreetmap website.

Anyways… happy mapping!

Location: Kampung Ampangan, Kuala Selangor, Selangor, Malaysia

Never ending edits

Posted by kucai on 10 August 2020 in English.

I believe I cleared most of east coasts’ osmose warnings a while ago. However, new warnings/errors keep popping up. Either osmose revised their rules, or new (non-local) mappers keep adding stuff that trigger osmose’s detection algorithm.

Also these mappers keep re-aligning existing ways to Maxar, despite there’s no indication that the new imagery is any more accurate than Bing. When 90% of the area was traced to Bing, why shift a portion to Maxar? If that’s the intention, why not shift everything to Maxar while you are at it and make them standardized. It’ll surely do wonders to your KPI, in addition to your ‘smoothing’ of highways.

That and reclassifying highways in a country you never step foot on.

Kratom mixed with Coke sounds good now…

-peace out-

Location: Kampung Santong, Dungun, Terengganu, 23100, Malaysia

Survey points etc

Posted by kucai on 30 June 2020 in English.

I have been registering surveyed points from the RTK gps units that I used at work. Usually, I updated those references to JOSM’s image offset database, used to be both for Bing and Maxar. Lately however, I just uploaded data for Bing only.

Part of the reason is that I found that Maxar has quite a lot skewy images despite their recentness. Their images are also at a different zoom level than Bing, so providing offsets for both Maxar and Bing seems to be pointless. Older Bings were nicer, but newer ones seems to be the same image used by Esri and Maxar, so they are equally good (or bad).

To overcome the different zoom scale of the imageries, I always defaulted to the ‘corrected’ Bing. In areas where new roads/developments can be seen only in Maxar, I will offset Maxar to Bing. I’d find a house closeby, trace it in Bing, and then move Maxar imagery to match. Hopefully, everything will remain consistent. Well, at least until new (paid/unpaid) mappers started tracing and aligning everything to Maxar that is..

I am still undecided regarding gps tracks. Quite a few issues: bad signal (resulting bad accuracy), kalmann filter, not enough traces covering x-y axis. Better than nothing though. I’d rather prefer a 10 minute averaging (or median value) gps position.

Anyways….

Dual freq GPS and map alignment

Posted by kucai on 29 July 2018 in English.

With the introduction of dual freq chip from Broadcom, and the maturing firmware on Mi 8, the possibility of 30cm accuracy becomes much closer to mappers (hopefully standalone receivers are in development too). How are you guys planning to deal with this? Do you think there will there be a map alignment wars? Will iD incorporate that imagery offset db plugin like in JOSM? Do we go nuts with different imagery having different lens zoom ratio on the same zoom layer and having align both to a unified offset?

The OCD in me wants to align everything precisely, but I sure appreciate the Soviet’s approach of “good enough”.

Much welcomed DigitalGlobe imagery

Posted by kucai on 12 June 2017 in English.

I have just explored the newly available Digitalglobe imagery around Malaysia, and was blown away with its coverage after all these years waiting for Bing. It was almost orgasmic! Not to mention the few that I worked with aligned almost exactly with my gps tracks, right in the middle of the road. Very happy, despite finding some sort of distortion (wavy terrace houses) with the imagery. Perhaps that could very well be caused by the unserviced surveyor’s theodolite during the housing area survey & marking stage.

Then it dawned on me..all those areas to be traced. That’s a seriously a huge amount of work to be handled by a handful of OSMers around here. Going to need a couple hours dedicated to tracing everyday whenever possible, starting with the important features first : streets in cities and towns, then following up with residential areas and the like.

Nevertheless, thank you Digitalglobe!

p/s anybody had trouble with JOSM (ver12039 here) acting up? At one point, text input doesn’t work anymore, but mouse actions work (pressing buttons & stuff). Just wondering although my Opensuse Leap 42.2 laptop isn’t that top notch either - btrfs, kde plasma & memory are possible issues. Maybe it is my laptop… anyways…. peace out.

Mapping lake shore

Posted by kucai on 7 August 2016 in English.

I started refining the shore boundary of Lake Kenyir, reputedly the largest (or 2nd?, not sure, another lake is being filled up in Sarawak) in South east asia. Soon enough, I realized that there are jillions of coves that might need to be mapped, which brings the question of what’s the acceptable level of fidelity when dealing with this kind of mapping. (this would certainly the case when mapping a country’s shorelines too.)

I finally ended up using the ‘triangle’ method. 2 points at the jutting land ends of a cove (which is also called a point in bass fishing:) ) and 1 in the middle of the cove. If that is still too rough, add a point to either side of the middle point, which usually is good enough to show up as a cove in higher zoom renderings.

Oh well.. I probably need to visit these places with fishing gear.. you know.. verification purposes and all that. hah!

Location: Hulu Terengganu, Terengganu, Malaysia

Aligning hiway to Bing imagery

Posted by kucai on 30 December 2015 in English.

Just finished aligning the northern half of Jabor-Jerangau highway (about 50km) to newly available Bing imagery. Since the highway was originally created based on a gps track (I think), there was quite an offset between the gps track and Bing.

Also in some areas, it seemed that hiways on Mapbox imagery lay much closer to accumulated gps track than Bing, but that can possibly be attributed to the difference between coarse level 17 zoom to Bings level 18/19 color imagery. Bing is about 20-30 meters away from gps tracks.

I think I’ll wait for more gps tracks before I offset the imagery. For now, I’ll focus on making pretty highways…

Thank you Bing.. need moarrr…

Location: FELDA Jerangau, Dungun, Terengganu, Malaysia

New sat image and alignments

Posted by kucai on 3 September 2015 in English.

Partially finished realigning buildings in Kota Bharu, Malaysia. It seems that a new sat image was available, and a mapper(s) traced the highways according to the new image but leaves the outline of buildings from older image intact. Therefore, those buildings have to be moved manually by selecting each one of them to align with the new image. (ignore lat long accuracy for the moment).

Perhaps, anyone who discovers an area with new sat image should first align existing hiways/buildings. Once the hiways/buildings matched with the new sat image, only then the sat image & hiways & buildings be offset to the accurate position. (can also be done the other way around, align the image to the existing highways/buildings, whichever is more accurate). That way we don’t have to align things two or three times manually. (Anyway, anybody knows how big an area is offset when we realign a sat image? One tile, 10sqkm, whole state?)

Or perhaps, people should try not to accidentally realign the map (new mappers especially).

Central Kuala Lumpur

Posted by kucai on 23 December 2014 in English.

I had to take the bus to the city center and back yesterday. While searching for the correct boarding locations (3 different locations within a kmsq), I noticed quite a lot has changed. During the wait for the arrival of the return bus, I wished that my ver 1 galaxy tab has the processing power to run vespucci to do on the spot corrections. Vespucci used to run & edit, but nowadays it runs without showing maps & data from osm anymore, only a black screen. Need to start saving for upgrade I guess (or win a lucky draw somewhere).

BTW, any of you visiting my country are more than welcomed to help map that area. :) (our currency is devaluing right now, more bang for your money!)

Location: Bukit Petaling, Kuala Lumpur, 50460, Malaysia