Enhancing workplace safety with artificial intelligence

For people who work mostly in the comfort of air-conditioned offices, workplace safety usually isn’t something to fret about. But for people like Home Team officers who work in dynamic environments that are filled with equipment, vehicles, machinery and weapon systems, workplace safety is no joking matter.

This is exactly why HTX engineers have leveraged artificial intelligence (AI) to create SafeX, an AI system that can quickly notify supervisors if it detects dangerous situations and safety lapses.

Developed by a team of engineers from HTX’s Platform Systems Capability Sustainment (CS) and Q Team Centre of Expertise (CoE), SafeX uses an AI Vision model to analyse CCTV footage and detect safety infringements, including high-risk work activities, dangerous vehicular movement, housekeeping lapses, and even personnel who aren’t wearing Personal Protective Equipment (PPE) in hazardous environments.

Upon detecting such incidents, SafeX immediately sends an alert to the workplace supervisor so that the safety lapse can be quickly remedied.

The team trained SafeX’s AI Vision model using a combination of open-source datasets for safety equipment and workshop machinery, as well as data from the marine workshop at the Police Coast Guard’s base in Pulau Brani.

The team selected the YOLO convolutional neural network (CNN) architecture as it is highly adept at object detection and classification. They then trained an ensemble of models to recognise various objects of different sizes. Following that, the outputs from the models were used to develop an algorithm that programmatically identifies specific activities and enforces safety standards within the shipyard environment.

The trial for SafeX has successfully concluded and the system is currently being explored for operational implementation.