Presence detection for lighting control with ML models using RADAR data in an indoor environment

Loading...
Thumbnail Image

URL

Journal Title

Journal ISSN

Volume Title

Perustieteiden korkeakoulu | Master's thesis

Date

2021-10-18

Department

Major/Subject

Data Science

Mcode

SCI3095

Degree programme

Master's Programme in ICT Innovation

Language

en

Pages

104

Series

Abstract

There is an ever-rising demand for presence detection in intelligent buildings for reliable and real-time lighting control. RGB cameras have proven to be accurate for detecting human presence. However, RGB cameras pose privacy concerns and therefore cannot be used for indoor environments like offices. The solution which is commonly used in the lighting industry is thermal sensing using Passive Infrared sensors (PIR). PIRs are low-cost sensors and they have low power requirements and simple installation procedures. However, PIRs cannot detect stationary people and their output is bursty. Additionally, PIRs are prone to significant false positives which often leads to power wastage. Even though they are a sustainable solution for lighting control, there is still a high scope for improvement. This thesis is concerned with lighting control using Frequency Modulated Continuous Wave (FMCW) RADAR, which tracks the human presence in the architectural premise. Unlike PIR, the range of the RADAR is long, and one RADAR acts as an alternative to numerous PIRs. Moreover, the RADAR data is large in volume, thus more reliable. The objective of this thesis is to measure true presence using RADARs. Deep Learning classification models are trained using RADAR data to make the machine understand the human presence. This thesis includes data collection from the RADAR, data labeling for the classification models, data visualization and exploratory analysis, and eventually data modeling using Deep Learning. The aim is to develop a generalized model for presence detection using RADAR data for lighting control. The proposed method achieves an accuracy of 98.6\% when predicting the presence. These results improve over the state-of-the-art accuracy in presence detection.

Description

Supervisor

Solin, Arno

Thesis advisor

Nasir, Omar

Keywords

FMCW RADAR, lighting control, machine learning, presence detection, deep learning, sensor control

Other note

Citation