CMU 18-848D: Embedded Systems for Internet of Things

Automated Retail Inventory Tracking

Developing a network of low-cost, embedded BLE beacon camera modules distributed throughout shelves in a retail store to achieve better inventory coverage at faster update frequencies and lower costs than all current solutions. Each 5"x3.5" module is equipped with internal WiFI, Bluetooth Low-Energy (BLE), a 5 MP camera, and a Raspberry Pi linux-based computer. The housing is design to contain all electronic parts and eventually be water-proof (IP68). A custom adjustable bracket is designed to mount universally to different variations of store shelves. The camera captures an image every minute or two (image capture frequency is an input) and uploads the image to the cloud using WiFi. A server then scrapes the uploaded image and performs a series of vision-based computations as outlined in the document below to determine quantity and stock status of every shelved item in retail stores.

Vision Algorithm for Inventory Tracking