Please use this identifier to cite or link to this item:
|Title:||Real-time helmet wearing determination AI algorithm: construction workplace health and safety||Authors:||Lee, Veatrice Wei Ling||Keywords:||Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence||Issue Date:||2022||Publisher:||Nanyang Technological University||Source:||Lee, V. W. L. (2022). Real-time helmet wearing determination AI algorithm: construction workplace health and safety. Final Year Project (FYP), Nanyang Technological University, Singapore. https://hdl.handle.net/10356/159171||Project:||C135||Abstract:||Singapore's Workplace Safety and Health (WSH) 2028 vision seeks to leverage on emerging technologies to become a world leader in the field of worker's safety. To realise that vision, a primary factor contributing to high risk at work needs to be tackled. In the construction industry, this factor would be the improper and non-use of personal protective equipment (PPE). Majority of previous research proposed the use of surveillance systems, using cameras to perform image processing to monitor PPE usage. However, blind spots make it difficult to monitor every angle of the work site. An alternative method suggested the use of micro-electromechanical system, which relies on sensors to collect the relevant data to determine is the PPE is donned through use of Logistic Regression (LR). In this study, the MEMS method is undertaken where the PPE in question is the safety helmet. The aim is to expand on the previous research by studying the performance of 3 other machine learning (ML) models on top of LR. Namely, Decision Tree (DT), Support Vector Machine (SVM), and Gaussian Naive Bayes (GNB). Additionally, the real-time capability of the prototype is also evaluated to gauge the solution's feasibility. Internal humidity (i.e., inside of helmet) and external humidity (i.e., ambient humidity) data is used to produce the parameters for the helmet wearing determining program: (a)ambient-microclimate humidity difference (AMHD), and (b)ambient-microclimate humidity difference rate of change (AMHDROC). The findings show that DT model performs similarly to LR model for goodness of fit with average median of 92.5%, average IQR range of 3.5%, and IQR difference of 3%. It is also superior in terms of lag time with average median of 4 ticks, average IQR range of 3.5 ticks, and IQR difference of 1 tick. This would show that DT model can be considered as an alternative to LR model as seen in its performance metrics. In turn, the underlying concept of this solution can be implemented for other PPEs to enable greater visibility for site managers with regards to PPE monitoring to aid in workplace safety practices.||URI:||https://hdl.handle.net/10356/159171||Fulltext Permission:||restricted||Fulltext Availability:||With Fulltext|
|Appears in Collections:||MAE Student Reports (FYP/IA/PA/PI)|
Updated on Dec 1, 2022
Updated on Dec 1, 2022
Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.