Deep learning-based construction activity classification

Abstract

Construction activities often produce excessive and prolonged vibrations that can be detrimental to adjacent infrastructure, equipment, as well as people. To mitigate the negative effects of construction-induced vibrations, vibration monitoring is usually implemented to analyse the impacts of the vibrations. However, these vibration data collected were mostly not fully utilised due to the lack of information such as labelling of data. Therefore, this study aims to collect vibration data on various construction activities followed by developing a deep learning (DL) algorithm to recognise the different construction activities. The classification of construction activity was performed by adopting a convolutional neural network.