How Accurate Is Korea's Public CCTV Dataset? What I Found Building a Nationwide CCTV Map
TL;DR
- Korea's published CCTV count is not the real camera count: private and undisclosed cameras are missing, while many records share one coordinate and some lists were uploaded more than once, so the row count exceeds the number of real locations.
- The raw 381,036 rows add up to 700,548 cameras, and 55,704 coordinates are registered more than once.
- 42.6% of install dates are blank, and placeholder coordinates like 36.0/127.0 and install dates from 1899 show up.
- Base dates range from 2020 to 2026, so cite the numbers together with their scope and base date.
While building Viewcone, a public CCTV map, I cleaned the entire national CCTV standard dataset (전국CCTV표준데이터) from Korea's Ministry of the Interior and Safety. Working through it answered a question naturally: does the published CCTV count equal the real number of cameras? No.
The dataset
| Item | Detail |
|---|---|
| Dataset | 전국CCTV표준데이터 (National CCTV Standard Data), Ministry of the Interior and Safety |
| Format | CSV, CP949 (extended EUC-KR) encoding |
| Total records | 381,036 |
| Key columns | management ID, agency name, address, purpose, camera count, install year-month, WGS84 lat/lon, base date |
| Providers | 228 municipality codes, 1,835 agency names |
This file is not a single national snapshot. It's a union of records each municipality and agency uploads on its own schedule, and that difference explains every problem below.
What I found
1. One row is not one camera
Each row is an installation point. A separate column holds the camera count, and summing it gives 700,548 cameras across 381,036 rows. The camera total is 84% higher than the row count. Multiple cameras often share one pole; the maximum camera count recorded in a single row was 201. When a headline says "Korea has X hundred thousand CCTVs," the number can differ by almost 2x depending on which figure is used. I compared the government's national totals beyond this dataset in how many CCTV cameras South Korea has.
2. 55,704 coordinates appear more than once
Grouping rows by lat/lon rounded to 6 decimals turns up 55,704 coordinates used by at least two rows (182,569 records in total, nearly half the file). Part of this comes from the repeated uploads in section 8: when an agency uploads its list again, each upload adds the same coordinate once more. Leaving those earlier uploads out still leaves 48,177 shared coordinates (149,284 records), and 47,272 of them (98%) are rows within a single upload, each with a different management ID. One Seoul alley, for instance, has six consecutive IDs in one upload sharing the same address, coordinates and install month. They look like separate installations or cameras at one address, each recorded as its own row but given one representative coordinate. The coordinates were probably geocoded at the address level or copied.
So Viewcone merges rows sharing a rounded coordinate, normalized purpose category, and agency into a single point (they'd land on the same pixel anyway) but sums their camera counts. Merging removed 95,231 rows, leaving 253,980 map points / 640,238 cameras. The math checks out: 700,548 − 83 (cameras on the 26 out-of-range rows) − 2,903 (cameras on the 2,576 placeholder rows below) − 57,324 (cameras on the 29,223 rows of earlier uploads in section 8) = 640,238.
Keeping only the first record of each shared coordinate would throw away the camera counts of separate installations and undercount the cameras. A list that appears several times has the opposite effect: summing counts the same cameras more than once. That is why the earlier uploads are left out before anything is merged.
3. Dummy coordinates pointing at two inland spots
1,579 rows sit at exactly 36.00000, 127.00000; another 990 sit at 36.1/127.1. Values this round strongly suggest placeholders rather than measured coordinates. 2,576 rows have both latitude and longitude rounded to one decimal place or fewer. They fall inside the configured bounds, so at first they passed the filter and thousands of cameras piled onto two points on the map, one in Iksan, Jeonbuk (36.0/127.0) and one in Nonsan, South Chungcheong (36.1/127.1). They skewed the regional counts too: Gongju City Hall's 1,579 cameras were counted in Iksan and Daedeok-gu's 1,317 in Nonsan, on top of those cities' own cameras (2,973 in Iksan and 2,942 in Nonsan on today's map). Since October 14, 2026, coordinates with both values at one decimal place or fewer are treated as unlocated and left out of the map and the regional counts.
The screenshot below shows the 36.00000, 127.00000 pile before that change: all 1,579 rows there list Gongju City Hall as the managing agency, yet they sat in farmland about 50 km south of central Gongju.

Beyond that: 26 rows fall outside the configured bounding box (lat 32-39.5, lon 123-132).
4. 42% of install dates are blank - and some say 1899
The install year-month field is empty in 162,131 rows (42.6%). About 2,200 contain only a year ("2026"), and 46 read 189912 or 190001, consistent with spreadsheet artifacts, like a default date near January 1900 leaking through a date-to-number conversion. Another 59 start with 1905, which that explanation does not cover.
5. Dozens of spellings for the same thing
The shooting-direction field is empty in 48% of rows (181,932), and 55 distinct filled values contain "360", such as "360도 전방면", "360도", "360도 전방", "360도전방면", "360" and "360°". Retention days show 30 days for 80.8% of rows but range from 0 to 180 days, with 61,788 blanks.
6. Each agency's data lives in a different month
The base date column spans January 2020 to September 2026. Seoul's most common base month is 2026-05 while Gangwon's is 2024-07, a 22-month gap, so "national CCTV status" is really snapshots spread across roughly six years and eight months. The full breakdown for all 16 regions is in How fresh is Korea's public CCTV data?
7. Agencies write coordinates to different precision
How many decimal places an agency writes also varies. Three decimals put a point on a grid of about 100 m, two decimals about 1 km, so a dot on the map can be that far from the camera. Incheon's Michuhol-gu office (인천광역시 미추홀구청) wrote all 3,176 of its rows with three decimals of latitude and two of longitude (for example 37.461, 126.64). Those rows pile up on a few points of a 0.01 degree grid: of Michuhol-gu's 314 map sites, 303 have coarse coordinates and 302 of them are on the 1 km grid.
Nationwide, 17,595 of the map's 253,980 sites (6.9%) have coordinates published only to about 100 m or 1 km, 1,654 of them to about 1 km. Decimals are counted as written in the source file, on whichever of latitude and longitude is coarser. I looked at geocoding the address on the same row instead, but found no address API I could use; the geocoding API comparison explains why.
8. Some agencies upload the same list more than once
Some managing agencies uploaded their whole list again under a new base date, and the source file still holds the earlier upload as well. Daegu's Seo-gu office (대구광역시 서구청) has lists of 1,000, 1,300, 1,350 and 1,599 rows all in the file, and each upload carries new management IDs, so by ID alone they look like different cameras. Since 2026-10-19 the map leaves out an agency's earlier upload when at least 80% of its rows reappear in the same agency's later uploads (same address or same coordinates) and the later uploads have at least 80% as many rows. That left out 38 uploads with 29,223 rows (57,324 cameras), and Daegu Seo-gu went from 1,060 sites and 5,246 cameras to 936 and 1,593. Finding repeated uploads in Korea's CCTV open data explains how I found the problem and how I set and checked the rule.
Why this happens
The standard data program works like this: the ministry defines the schema, each agency fills it in. Across 228 municipal codes, the staff, update cycles, input habits and ways of uploading a new list all differ. The format is shared but the interpretation isn't - whether "one point" means one camera or one pole varies by agency. Coverage varies too: private CCTV, cameras managed directly by police or military, and non-disclosed installs are simply absent.
What to keep in mind when reading the numbers
- Published count ≠ actual camera count. It undershoots by excluding private and undisclosed cameras, while many separate records share one coordinate and some lists were uploaded more than once, so the row count exceeds the number of distinct locations. The error runs both ways.
- Check base dates before comparing regions. Putting a 2024 snapshot next to a 2026 one bakes two years of new installs into the comparison.
- "Cameras per 100k residents" style metrics need the same coverage caveat. The regional breakdown shows how agencies' different registration and disclosure practices play out in the raw counts.
How Viewcone handles it
Before anything reaches the map, the pipeline:
- Drops anything outside the configured bounding box (lat 32-39.5, lon 123-132).
- Drops rows whose latitude and longitude both have one decimal place or fewer (an 11 km grid): such a row cannot mark a camera, so it stays off the map and out of the regional counts. 2,576 rows (2,903 cameras) such as 36.00000/127.00000 fall here.
- Leaves out an agency's earlier upload when at least 80% of its rows reappear in the same agency's later uploads and those hold at least 80% as many rows. 29,223 rows (57,324 cameras) fall here.
- Collapses rows with the same coordinate, normalized purpose category, and agency into one point, while summing their camera counts.
- Draws sites published only to about 100 m or 1 km as faded points with no coverage area, leaves them out of the check for whether your location is inside a camera's range, and does not geocode them again from their address.
- Maps "360도 전방" and friends to standard codes.
- Keeps the source and base date visible next to the numbers.
The map therefore shows cleaned points, not raw rows; per-region camera counts are on the region pages. The remaining error (absent private cameras, uneven agency coverage) can't be fixed until the source itself changes, so the map states that limitation explicitly.
When you read a number on Viewcone, check the base date shown next to it.
FAQ
Are there cameras missing from this dataset? Yes - privately installed CCTV (shops, apartment complexes) and cameras withheld for security reasons aren't included.
Why does Viewcone show fewer points than the raw file? Rows sharing a coordinate, purpose category, and agency occupy the same pixel on the map, so they collapse into one point - but their camera counts are summed, so the total only loses the 83 cameras on out-of-range rows, the 2,903 on placeholder coordinates and the 57,324 in earlier uploads of a list that was uploaded again.
How fresh is the data? It depends on the agency - base dates inside a single file range from 2020 to 2026.
Source
- 전국CCTV표준데이터 - Public Data Portal (Ministry of the Interior and Safety)
- Analysis uses a file downloaded 2026-09-20