How Fresh Is Korea's Public CCTV Data? Comparing Record Ages Across 16 Regions
TL;DR
- Korea's national CCTV standard dataset is not one snapshot but a union of records refreshed on different schedules, with base dates from 2020 to 2026.
- The median record age is 4 months in ten regions including Seoul and Busan, 5 months in Daejeon and 26 months in Gangwon.
- 58% of Gangwon's and 34% of Daejeon's records are over two years old, so compare regions only alongside their base dates.
In the previous post I showed that the Ministry of the Interior and Safety's national CCTV standard dataset (전국CCTV표준데이터) mixes base dates from 2020 to 2026 in a single file. The base date tells you which point in time a record describes and is separate from when the camera was installed. It is closer to when the managing agency last refreshed the data. The dataset's name suggests one national snapshot, but the file is a union of snapshots taken at different times.
To see how far apart the regions' base dates are, I assigned 346,541 records to regions by their coordinates and compared their base dates.
Dataset and method
| Item | Detail |
|---|---|
| Dataset | 전국CCTV표준데이터 (National CCTV Standard Data), Ministry of the Interior and Safety |
| Snapshot date | 2026-09-20 (download date) |
| Total records | 381,036 |
| Analyzed | 381,010 rows inside the configured bounding box (lat 32-39.5, lon 123-132), minus 2,576 placeholder coordinates and 29,223 rows in earlier uploads of re-uploaded lists: 349,211 |
| Region assignment | Region polygon containing each record's coordinates |
| Age | snapshot date − base date in months |
2,576 records with placeholder coordinates given to one decimal place or fewer (such as 36.0/127.0) carry no usable location and were dropped. Where an agency uploaded its whole list again under a new base date, the 29,223 records in earlier uploads that the later list largely repeats were dropped too; kept, they would count the same cameras again under an older base date (see how the repeated uploads were found and removed). Another 2,670 records (~0.8% - island and borderline coordinates) fell outside every region polygon and were excluded, leaving 346,541 records across 16 regions. Gwangju and Jeonnam are aggregated under the merged Jeonnam-Gwangju boundary.
Region × base-month timeline
Rows are the 16 regions, columns are base months. Cell color shows the share of that region's records carrying that base month - shares within each region, not absolute counts, so high-volume regions like Seoul or Gyeonggi don't automatically look darker.
- Most regions have their brightest cell in May 2026. Of the 346,541 records analyzed, 153,875 (44%) carry a May 2026 base date; across the whole file, the busiest single day is 2026-05-01 with 21,746 records. Many agencies appear to have refreshed around the same period.
- Gangwon's big block sits at 2024-07: that one month holds 43% of its records, and 42% date from September 2024 or later.
- Daejeon keeps two older blocks. May 2026 is its largest month (47%), but records based in October 2021 (14%) and June 2022 (20%) together make up about a third.
- Jeju is almost entirely two batches, 2025-05 and 2025-11, with barely any refresh since.
- Sejong is concentrated in a single month, May 2026: 1,678 of 1,679 records, which looks consistent with a wholesale refresh at one point in time.
Median record age by region
For each region I computed every record's age in months, then took the median. Median over mean because a handful of 2020-era rows would drag the average around.
| Region | Records | Median | Mean | Newest | Oldest |
|---|---|---|---|---|---|
| Seoul | 55,387 | 4 mo | 9.0 mo | 2026-07 | 2021-04 |
| Gyeongnam | 38,364 | 4 mo | 6.6 mo | 2026-09 | 2020-09 |
| Gyeongbuk | 32,987 | 4 mo | 12.5 mo | 2026-06 | 2021-10 |
| Chungnam | 29,047 | 4 mo | 13.2 mo | 2026-07 | 2021-10 |
| Busan | 17,095 | 4 mo | 8.9 mo | 2026-06 | 2024-01 |
| Daegu | 17,033 | 4 mo | 10.9 mo | 2026-06 | 2021-10 |
| Incheon | 13,852 | 4 mo | 6.5 mo | 2026-05 | 2021-07 |
| Chungbuk | 13,232 | 4 mo | 5.8 mo | 2026-07 | 2022-06 |
| Ulsan | 5,933 | 4 mo | 4.1 mo | 2026-06 | 2024-07 |
| Sejong | 1,679 | 4 mo | 4.0 mo | 2026-07 | 2026-05 |
| Daejeon | 4,224 | 5 mo | 21.2 mo | 2026-05 | 2021-10 |
| Gyeonggi | 53,699 | 9 mo | 14.3 mo | 2026-07 | 2020-01 |
| Jeonbuk | 13,625 | 9 mo | 9.7 mo | 2026-06 | 2022-06 |
| Jeonnam-Gwangju | 24,403 | 11 mo | 16.2 mo | 2026-06 | 2020-03 |
| Jeju | 1,012 | 16 mo | 13.9 mo | 2025-11 | 2024-07 |
| Gangwon | 24,969 | 26 mo | 20.7 mo | 2026-07 | 2021-07 |
- Ten regions sit at a 4-month median: Seoul, Gyeongnam, Gyeongbuk, Chungnam, Busan, Daegu, Incheon, Chungbuk, Ulsan, Sejong. Over half their records are May 2026 or newer.
- Gangwon sits at 26 months, more than six times Seoul's median.
- The gap between median and mean carries real information. Seoul's median is 4 months but the mean is 9.0: mostly fresh records with a long older tail. Daejeon (5 vs 21.2) shows an even wider gap.
- Jeju's newest record is 2025-11, so the freshest thing in this snapshot is already ~10 months old. No refresh appears to have happened in nearly a year.
- Sejong's mean equals its median (4.0/4), consistent with a single-batch update.
What "stale" actually means - freshness buckets
A plainer measure is the share of each region's records that are over two years old.
- 58% of Gangwon's records are over 2 years old. More than half its data predates September 2024. In Daejeon the share is 34%, in Jeonnam-Gwangju 31%.
- Seoul still has 7.5% over 2 years: fresh on median, but about 4,100 records date from August 2024 or earlier.
- Busan is bimodal: 73% under 6 months and 20% over 2 years.
- Gyeonggi's middle band is thick (40% at 6-12 months): a substantial share of records hasn't moved to the newest base dates.
Why regions differ this much
The structure of the dataset makes it inevitable. The ministry standardizes only the format; over 200 agencies fill in and upload their own content. There's no unified rule for when to refresh.
So each agency ends up on its own cadence. Some regions look like they were refreshed in semiannual sweeps; others still hold large shares of 2024-era records. Cadences can differ even within one region across cities, counties and departments. The "fresh core plus older tail" pattern in Seoul and Busan looks consistent with that kind of mixture.
What to keep in mind when reading the stats
- Compare camera counts only alongside base dates. Putting 2024-based and 2026-based regions side by side lets up to two years of installs and removals distort the comparison. "Gyeonggi has more cameras than Gangwon" may partly be a freshness gap.
- A stale base date doesn't mean the camera is gone. The camera may exist while the data was simply never refreshed. Read an old base date as "not confirmed recently," not as "absent."
- Mixtures exist within each region too. A regional average hides agency-level gaps: one county can be current while a neighboring district sits two years behind.
The same limitation shows up on the map
Viewcone inherits this data. Region pages and the map carry source and base-date labels, but a large share of Gangwon's points may still reflect 2024 positions while Seoul reflects 2026.
FAQ
Does an old base date mean the CCTV is gone? No. The base date is when the data was last compiled, which is separate from whether a camera was actually installed or removed. An old base date means the record was last checked a long time ago.
Why is Gangwon's data so old? Each managing agency refreshes on its own schedule. The standard unifies the format, not the refresh cycle. Some agencies' data appears to have gone unrefreshed for a long time.
How accurate is the region assignment? Records were placed by the region boundary containing their coordinates - camera location, not agency address. 2,670 boundary/island records (0.8%) matched no polygon and were excluded, and 2,576 placeholder coordinates (one decimal place or fewer) and 26 rows outside the coordinate bounds were dropped as unlocated.
Will the next snapshot look different? It may. When an agency refreshes its records, that region's distribution shifts toward recent months. Every number here comes from the snapshot downloaded 2026-09-20.
Source
- 전국CCTV표준데이터 - Korea public data portal (Ministry of the Interior and Safety)
- The analyzed file was downloaded 2026-09-20