Elevator occupancy counter camera

For my client Armando, I implemented an elevator people counter camera. He is an elevator installer and works with industrial partners. One of them had a freight elevator that needed to double check an operator is present before activating.

Key issues

Armando needs to pass the cost of the device and development to his own clients, so he wants a solution that is affordable. Moreover, he needs to have full control over the output of the detection camera so that it can be easily integrated into his existing setup and workflow. Sometimes, an HIGH signal on a digital pin is all he need to control an actuator. Other times, he needs the exact number of detected people over RS485. 

Problem statement

This project is about deploying a cheap camera device that monitors h24 his field of view and counts how many people are there. It is intended to be deployed inside elevators, but nothing limits its applicability to other settings.

To stay privacy-compliant, the detection algorithm runs locally on the device and never reaches the internet. When one or more people are detected, a custom output is generated from the camera.

How To

A local deep learning model is flashed into the camera firmware to detect people accurately and in realtime. The camera collects the number of detected people and can be configured to produce a custom output as per project requirements: pin toggle, RS485, Serial output, I2C... The detection is so fast that is takes 10ms to process a single frame. Special image analysis pre-processing is possible to accomodate low-light, noisy environments with little impact on accuracy. 

The camera has even WiFi connectivity, so that it can be connected to a monitoring server for integration into larger systems.

How much does it cost?

Apart from my development cost, such AI-capable camera costs around 25$.