Solutions by typeMobility analysis
How many times has it nearly happened?
A near miss leaves no trace
Ten to five, the street outside the school. A child with a backpack steps out from between two parked cars and runs across the middle of the road, away from the crossing. Another car is coming down that lane and brakes in time. Nothing happens. The driver moves on, the child reaches the sidewalk, and not one line is written in the day’s log, because nothing happened.
11 cameras across 7 sectors
This is how you ask for it
«Alert me when a child steps into the road away from the crossing, and keep how often it happens, at what time and where.»
The same thing happens every afternoon, at the same hour and the same spot, and not one line is written about it. The day a car doesn’t brake in time, somebody will ask whether this had happened before. It will have happened a hundred and twenty times, and nobody was counting. Crossings get moved or signalized with data, and that data is what is missing.
You decide once
The city, or the factory, decides once which surface is roadway and which isn’t, in what hours it matters and what happens when it occurs: whether it rings in the control room, whether it’s only counted, or both. That second part is the one almost nobody asks for and the one that pays best: that it’s always logged, even when nobody had to brake. IRIS applies it every day at the same hour, without anyone remembering to switch it on.
The school street is the version everybody recognizes, but the same scene repeats inside a warehouse, and there the vehicle weighs four tons. An operator walks along the edge of the green walkway with the pedestrian symbol behind him, while a forklift moves up the vehicle lane beside him. At the aisle 4 intersection another one pulls out with a pallet of boxes stacked above the driver’s head: straight ahead he sees nothing, and the person on foot jumps aside. In the engine aisle, an AGV arrives from the left just as a forklift crosses from the right.
In all three places there are two things to ask of the camera, and they differ. The first is the alert now: a person and a vehicle have coincided in the same stretch, with the time and the video kept so it can be watched. The second is the count: how many times it happened this week, at which intersection, at what hour and with which machine. The first is for the conversation with the driver. The second is for moving a rack, backing up a traffic direction or fitting a mirror, which is what removes the risk.
And it’s worth saying what this isn’t: a judgment on anyone. It’s rarely one person being careless; it’s a badly drawn aisle or a load blocking the view. With the data in hand, the talk is about what gets changed on the floor.
What changes
Near misses used to vanish. If nobody fell, there was no report; with no report, it never existed. And changes to the street or the aisle were decided from what people remembered in a meeting.
Now every encounter is logged with time, place and video, and at the end of the month there is a list. You can see the intersection that holds half the cases and the hour when almost all of them cluster. That’s what moves a mirror, a barrier or a schedule.
Where we have seen it
CAM-236A pedestrian jumps back at the crossing
«Tell me when a person and a vehicle meet at the crossing, and save me the video.»
CAM-12Vehicle parked on the sidewalk
«Tell me if a vehicle parks on the sidewalk, and how long it has been there.»
CAM-254Only a white line separates the pedestrian from the forklift
«Tell me when a forklift and a person are in the same stretch of aisle at once, and count how often it happens each day.»
CAM-271The pedestrian jumps clear at the crossing
«Tell me when a person and a forklift are in the aisle 4 crossing at the same time, and save me the video.»
CAM-470A car reaches the crossing just as someone is on it
«Alert me when a car enters the crossing while somebody is walking across it.»
CAM-63People walking on the roadway
«Tell me right away if you see anyone walking on the roadway, and all the more so near the tunnel.»
CAM-140A child crosses the road alone
«Tell me when a child steps into the road away from the crossing, and keep me a count of how often it happens, at what hour and at which spot.»
CAM-41Boulevard: cars, bikes and people
«Also tell me how many people ride past without a hard hat, so I know whether we need a road-safety campaign.»
CAM-272Crossing the walkway without seeing what is ahead
«Tell me when a forklift crosses the walkway too fast or with the load blocking its view.»
What it doesn’t do
IRIS does not brake the car or stop the forklift: it alerts, and a person decides. And it does not know who the pedestrian is: it identifies nobody and keeps no record of people.
Also, a coincidence in the image is not always a real risk: two things that look adjacent can be fifteen feet apart. The criterion is tuned on site, with the people who know the aisle.
Questions
- Does it tell a child from an adult?
- Not by name, and it doesn’t need to. IRIS identifies nobody: it does not recognize faces and does not know whose child is crossing. What it sees is a small person in a place where cars go, and that’s what triggers the rule. For what is needed here — somebody looking at that spot now, and the event counted for later — knowing who they’re adds nothing, and storing it would carry a cost not worth paying.
- Can it stop the forklift or the AGV?
- Not on its own, and that boundary should be respected. Stopping a machine is a safety function of the installation, with its own components, wiring and certification; a camera can’t take that place and shouldn’t claim to. What IRIS does is alert at once, keep the video and count how often it happens. If the site has a stopping system, that system rules; IRIS contributes the eye that sees the whole intersection.
- Does it also count the times when nothing happens?
- Yes, and it’s the most valuable thing it does here. A collision is data that arrives late: by the time it exists there is nothing left to prevent. Near misses, on the other hand, happen every week and are recorded nowhere today. Counting them turns a hunch — “that intersection is dangerous” — into a list with dates, times and places. That’s what justifies a mirror, a reversed direction or a different delivery schedule, and it justifies it to whoever pays.
- And at night or in the rain?
- It works, but not equally, and that should be said. A person in dark clothes, at night, with the road wet and reflecting headlights, is the hardest situation there is for any camera — and for the driver too. On a street with lamps it holds up well; on an unlit stretch you need lighting or a camera built for it. Inside a warehouse or a parking lot the problem doesn’t arise, because the light is always the same.
Every simulation
Something has spilled
Tell me if there is liquid or foam outside its place and it is still there two minutes later.
- The sentence
- Where it applies
- What it finds
- The same rule
Not seeing yours here?
These words are the ones we have built, not the ones IRIS understands. The list falls short on purpose: you ask for it in a sentence, and sentences don’t run out. Tell us what is in front of your cameras and we’ll tell you whether it understands it, or not yet.
And the best part
Finding something in yesterday’s videoshould take as long as asking for it.
No more scrubbing back through six hours with your finger on the mouse. You type what you’re after, in plain words, and the matching moments come up. Same cameras as always.
Tell us your problem and we’ll say whether IRIS understands it, or not yet.
A person answers, the same working day.