A learning model for automated construction site monitoring using ambient sounds

Construction site monitoring is an important task to analyze, measure, and monitor the activities in the construction site. In order to present/develop an automated construction site monitoring model, many machine learning methods have been presented in the literature. This work aims to develop an automated activity identification and construction vehicle classification model using sounds. Thus, two ambient sound datasets were collected. A new learning method is proposed to classify the collected sounds, and this model is named BTPNet21 since our proposal uses a binary and ternary pattern with a pooling function to extract features. Iterative neighborhood component analysis selector chooses the most significant features, and the support vector machine is utilized as a classifier. Our proposal attained 99.45% and 99.17% accuracy rates on the collected sound datasets consecutively. These results demonstrate that the success of the introduced BTPNet21 for sound-based automated construction site monitoring.