Optical Sensors

Small ceiling-mounted optical sensors measure transitions of persons across a threshold, while providing complete privacy protection.

BlueZoo is developing optical sensors to measure the transition of persons across thresholds, such as through doorways.  The sensors are privacy friendly and GDPR compliant, relaying no personal identifying information.

Every few seconds, the ceiling-mounted sensors report with 99% precision how many people have transitioned in each direction across the threshold.

These optical sensors measure transitions, people in and people out, and the speed of the transition (walking or running). By comparison, BlueZoo Wi-Fi sensors measure different details like dwell time, unique visitors, and other advanced metrics–albeit with only 90% precision. This makes BlueZoo optical sensors complementary to BlueZoo Wi-Fi sensors.

The high accuracy of optical sensors make it easy to measure precisely how many people are in a small space like an individual conference room or office. Measuring the occupancy of small spaces with low occupancy rates is not a good use case for Wi-Fi sensors, but is a good use case for optical sensors.

Optical sensors are possible due to recent major advances in semiconductor technology. Only since 2020 has it been possible to affordably execute complex machine learning models “at the edge”.  The benefits of edge computing are two-fold:

  • Privacy:  The only data uploaded from an optical sensors are numeric counts and aggregate. No personally identifiable information is ever captured or uploaded to cloud servers, or even stored at the camera. Images are processed at the sensor and then discarded.

  • Low bandwidth:  Optical sensors put exceptionally low load on internet connectivity, uploading less than 1 MB/hour per sensor, sending compressed data over an encrypted channel. These low data rates make cellular backhaul efficient and affordable.

The sensors use Google “Coral” boards that feature Google Tensor Processing Units (TPUs), designed expressly to execute neural network models “at the edge”. The boards are low power and do not require heatsinks with fans.

All sensors can be powered by a micro-USB connection (5V at 600mA). Some sensors are powered by power-over-ethernet (PoE) connections. PoE is an attractive solution for power delivery for low-power devices like our sensors because running ethernet cables (cat-5, cat-5e, or cat-6) is generally far less expensive than running line-voltage wires (120v or 240v) with their attendant safety requirements (e.g. conduit).

Get started today

Reach out to us — we’re here to help.

People counter sensor