Utilization of Artificial Intelligence (AI) to Monitor Construction Safety



Investigators


There has been a significant number of researches to reduce the number of injuries and fatalities resulting from non-compliance with PPE standards. To automate the monitoring of PPE standards, the systems could be categorized into two major types: 1) Sensor-based and; 2) Vision-based. However, both these methods do not provide real-time discoveries. Artificial Intelligence (AI) is being implemented in many different fields such as defense, security, healthcare, etc. to achieve the highest level of efficiency. In this study, the researchers have proposed the use of AI to monitor PPE compliance. Researchers have utilized the pre-built AI system based on Machine Learning (ML) and tested the model on BSCI service-learning classes.

Currently, to achieve real-time object detection, Artificial Intelligence/ Deep Learning/Machine Learning (AI/DL/ML) is being used in many different industries like defense, healthcare, safety, and security, etc. Particularly, Convolutional Neural Network (CNN) based approach; You-Only-Look-Once (YOLO) is being widely used for image classification and detection. YOLO is an object detection system targeted for real-time processing. It divides the input image into a 13×13 grid. Each grid cell is capable of predicting one object. For example, in the picture below, input was given to detect a bicycle, car, and a dog with different output colors to identify the differences.

RESULTS
The ML model was used on different images from the service-learning classes at BSCI. It can be seen in the figures below that the ML model has identified most of the hard hats/no hard hats on individuals on site. A detailed response can also be seen. However, in figure 5, it does not detect any hard hat on the person on the right-hand side. Therefore, some irregularities exist, which need to be resolved.

See more in:
CCIC, Research