Find A Seat App @ HKU Library

An app that tells you whether a library seat is ‘empty’, ‘hogged’, or ‘seated’ on a floor plan

Features

Seat Status

The app if a seat is completely empty, hogged with belongings on the table or chair, or seated.

Near-real time

The latency is not larger than 1 minute.

Easy set-up

The app provides interface for setting up on a new floor plan.

Methodology

Image Recognition

High-angle Image

Source images are taken from cameras mounted at the ceiling

Homography Transformation

Transform image to bird's eye view using OpenCV

Image Classification

Locate the presense of tables, chairs, persons, and other items using pre-trained deep learning models such as Yolo

Plot tables and chairs

Locate the location of tables and chairs and generate a 2D floor plan

Derive seat occupancy

Perform localization of persons and find out the seat occupancy information from position of persons and belongings in respect to the tables and chairs

Methodology

Internet of Things

Sensors

Suitable sensors are attached to the chairs. Potentially we have capacitive sensors, pressure sensors, IR detectors, temperature sensors and accelerometers in our wish list.

RSSI Indoor positioning

Either Bluetooth-based or Wifi-based RSSI will be adopted upon our investigation.

Data visualization (1): plotting a floorplan

Locate the location of tables and chairs and generate a 2D floor plan

Data visualization (2): deriving seat occupancy

Perform localization of persons and find out the seat occupancy information from position of persons and belongings in respect to the tables and chairs

Our Team

Lai Cheuk Wing

Year 4, BEng(Computer Science)

Wong Sun Day

Year 4, BEng(Computer Science)

Our Supervisor

Documentation

Project Plan

Interim Report

Final Report