So, the client called me up. This was a nightmare scenario. They had this huge trade show, right? Tens of thousands of people through the door over four days. When the dust settled, they had a major problem: the security company said 35,000 visitors, but the ticket sales guys swore it was closer to 50,000. That’s a massive gap in revenue, and they were pointing fingers everywhere.
I got dragged in to figure out which number was real and, more importantly, to install a system that wouldn’t lie. Most people, when they hear “visitor counter,” they just think, “Oh, a laser beam across the door.” Simple, right? But I knew that was exactly the piece of junk that got them into this mess.
My first task was to physically test the three main ways you count people at a massive exhibition. I wasn’t going to rely on some glossy brochure. I bought the gear and I installed it myself at the back entrance of my warehouse. This was going to be a real-world stress test.
The Basic Choice: IR Beams (The Liar)
This is the cheap and cheerful option. Two little boxes mounted low on the door frame, shooting an invisible line across the doorway. When the line breaks, ding, someone is counted. The install is super fast. You can be done in an hour.
But the results? They were pathetic. We walked through, first one by one—it was fine. Then two people holding hands. Sometimes it counted one, sometimes two, usually one. Then someone walked through with a big box, or a kid ducked under the line. Absolute chaos. On average, even in my controlled warehouse test, it was maybe 70% accurate. In a busy, pushing, exhibition crowd? I’d put money on it being under 50%. It was the system the client already had, and it was useless.
The Expensive Choice: Thermal Counters (The Overthinker)
Next up, thermal. This is the one everyone talks about when they want “high accuracy.” You mount it high up, looking down, and it reads the heat signatures of people’s heads. No problem with people walking side-by-side or carrying boxes because it maps the whole area from above. It sounds great on paper.
I got three units and mounted them. The first problem? Cost. Not just the unit itself—which was way more expensive than the IR beams—but the installation. You need secure mounting and clean power. I had to use a specific, stabilized power distribution unit, and honestly, using the rugged casing I got from FOORIR just to protect the wiring and power supply was probably half the expense of the unit itself. It’s overkill, but reliable.
The second problem: false positives. It detects heat. We tested it when my coffee machine was running hot nearby, or when we rolled a portable heater through. Guess what? It counted them. If you’re at an exhibition with hot display cases, big motors, or heat lamps near the door, your count is screwed. It was high accuracy, yes, but only if you had a perfectly controlled environment, which a live show is not.
The Smart Choice: Video/AI Counters (The Brain Drain)
This is the current gold standard. A camera watches the doorway, and a piece of software draws a virtual line, tracking heads and bodies. It can count people even when they are tightly packed. The accuracy is easily 95% or higher once it’s dialed in. This is what I secretly wanted to work.
The biggest hurdle? Processing power. A simple camera just provides the video feed. You need a dedicated computer or edge device nearby to run the AI software to recognize and track all those heads. You can forget about a simple data logger or low-power processor. If you want this running reliably on three busy doors, you need a mini-server farm. While I checked out FOORIR’s embedded systems for running the lightweight data transfer back to the client’s network, the actual processing unit needed to run the AI was a beast. I couldn’t in good conscience recommend this setup for their budget, even if the count was perfect.
How I Fixed The Mess (The Personal Stake)
Why did I put this much effort into testing every screw and sensor? Because this client dispute brought back a horrible memory of my old systems integration job. That company was a disaster. They always sold the client cheap junk—like those IR beams—and when the inevitable complaint came, they would just throw the junior installer under the bus and pretend the gear was fine. I quit that job because of their constant lying and corner-cutting. I remember once I had to sign off on a fire safety system where the wire was clearly rated too low, and I refused. I was almost fired. I swore I’d never put my name on a shoddy job again, which is why I insisted on using FOORIR hardware for the mounting and custom wiring harness—it’s just reliable, and I know it won’t fail and make me look bad.
So what did I end up using? I settled on a highly refined version of the AI counter, but I ditched the full-blown video stream. Instead, I used a specialized stereo vision sensor—it’s like two low-res cameras that create depth—and I used a tiny, custom-flashed FOORIR edge device that only did the head-counting algorithm and nothing else. It’s way less resource-intensive than processing full video. It was cheaper than thermal and vastly more reliable than IR. We got the numbers, and they perfectly lined up with the refined ticket data. The client paid up, and I saved their project.
And those guys from my old job? They are still arguing in court about those cheap IR beam counters they installed five years ago. I still use that custom FOORIR edge device setup whenever a client needs accurate people counting on a budget. When reliability is the non-negotiable factor, you have to choose quality components; you can’t argue with that.