How many security cameras a building needs, and where, depends on the places to be seen and on the shadows of shelves, walls and buildings. Here are ten facilities that an AI counted on the plan in earlier articles: shops, an office, apartment buildings, a day nursery, a warehouse and a car park. It used the tools it calls over MCP (suggest_cameras and get_visibility).
1. How it is counted
- The places to be seen, such as the doors, the checkouts and a cash machine, are given as areas, each with a weight. The share is the area seen, added up with those weights.
- The cameras are ordinary ones on walls or posts (90° across) and all-round ones on the ceiling. How far they see, how high they go, and where and which way they may point are each article's assumptions.
- From the places a camera may go, the tool picks the one that adds most, one camera at a time, and then swaps cameras to do better.
2. How many, and how much they see
| Facility | To be seen | Cameras | Number | Seen |
|---|---|---|---|---|
| A convenience store | 100.06 m² | Ordinary, on walls | 6 | 91.9% |
| A drugstore | 495.75 m² | Ordinary, on walls | 8 | 91.3% |
| A drugstore | 495.75 m² | All-round, on the ceiling | 8 | 92.9% |
| An office | 476.25 m² | All-round, on the ceiling | 8 | 96.6% |
| A 24-hour gym | 201.88 m² | Ordinary, on walls | 4 | 95.5% |
| A 24-hour gym | 201.88 m² | All-round, on the ceiling | 4 | 96.0% |
| A coin laundry | 76.81 m² | Ordinary, on walls | 1 | 85.6% |
| A coin laundry | 76.81 m² | Ordinary, on walls | 2 | 98.3% |
| A coin laundry | 76.81 m² | All-round, on the ceiling | 1 | 97.5% |
| An apartment entrance | 45.69 m² | Ordinary, on walls | 1 | 80.7% |
| An apartment entrance | 45.69 m² | Wide, on walls | 1 | 87.5% |
| An apartment entrance | 45.69 m² | Ordinary, on walls | 2 | 94.3% |
| An apartment entrance | 45.69 m² | Wide, on walls | 2 | 95.0% |
| An apartment entrance | 45.69 m² | Ordinary, on walls | 3 | 97.0% |
| A day nursery | 466.63 m² | Ordinary, on walls | 4 | 95.2% |
| A day nursery | 466.63 m² | Ordinary, on walls | 6 | 98.7% |
| A warehouse | 1,575 m² | Ordinary, on walls | 10 | 83.0% |
| A warehouse | 1,575 m² | Ordinary, on walls | 16 | 94.3% |
| A shop's car park | 1,293.25 m² | Ordinary, on walls | 4 | 88.4% |
| A shop's car park | 1,293.25 m² | On the wall and posts | 6 | 98.9% |
| An estate's grounds | 3,557 m² | On street-light posts | 8 | 41.1% |
| An estate's grounds | 3,557 m² | On outer walls and posts | 16 | 73.5% |
| An estate's grounds | 3,557 m² | On outer walls and posts | 23 | 84.8% |
- Each camera adds less. In the warehouse, ten cameras see 83.0% and sixteen see 94.3%. The first camera sees the most; each one after it adds less and less.
- In a small room, one all-round camera on the ceiling. In the coin laundry, one camera on the ceiling sees 97.5%; ordinary cameras on the walls see 85.6% with one and 98.3% with two.
- A wide lens helps when there is only one camera. At the apartment entrance one camera sees 87.5% instead of 80.7%; with two cameras the shares are 94.3% and 95.0%, hardly any different.
- Outdoors, distance and the shadows of buildings count. In the estate's grounds, eight cameras on the street-light posts see 41.1%, sixteen see 73.5% when the blocks' outer walls are used too, and twenty-three see 84.8%: one more by the refuse store and six on three new posts. In the shop's car park, four cameras on the shop wall see 88.4%, and six see 98.9% once two posts go up by the road.
3. Rooms that should not be seen
| Facility | Room | Doors open | Doors shut |
|---|---|---|---|
| An office | Toilets | 23.6% | 0% |
| A 24-hour gym | Women's changing room | 23.0% | 0% |
| A drugstore | Toilets | 0.4% | 0% |
| A day nursery | Formula and baby-bath room | 0.3% | 0% |
| A warehouse | Toilets | 0% | 0% |
- A camera need not point at a toilet or a changing room to see into it through an open door. In every facility, with the doors shut, 0% was seen.
- Put up a notice. The Q&A of Japan's Personal Information Protection Commission (Q1-13) says that a business that uses a security camera to take images in which a person can be identified should specify the purpose as far as it can and use the images only for it. Where people cannot easily tell that a camera is taking their picture, the business should take steps such as a "security camera in operation" notice.
4. The articles
- How many security cameras does a convenience store need?
- How many security cameras does a drugstore need?
- How many security cameras does an office need?
- How many security cameras does a 24-hour gym need?
- How many security cameras does a coin laundry need?
- How many security cameras does an apartment entrance need?
- How many security cameras does a day nursery need?
- How many security cameras does a warehouse need?
- How many security cameras for a shop's car park?
- How many security cameras does an apartment estate need?
5. How to ask an AI
I want security cameras in this shop (as drawn). Weighting the door and the checkouts, how many do I need, and where, so that 90% of what must be seen is seen? Check, too, that the toilet and the staff room are not seen.
6. What this does not check
- The shares are counted on the plan, by whether a line of sight is clear and within reach. Image quality, light, glare and whether faces can be made out enter only through the assumption of how far a camera sees.
- People blocking the cameras' view were counted only at the convenience store's lunchtime peak.
