RoboticsCats: Construction Site Fire Detection

Every year there are many fires on construction sites and in buildings undergoing refurbishment; people are injured, properties are destroyed, including irreplaceable heritage buildings and the industry suffers disruption and unexpected costs from which many never recover. Even if the construction project survives the fire, the experience will almost certainly demoralise the construction crew.

As a typical concrete jungle, Hong Kong has a high density of large residential and office buildings. Commonly, a building construction site is next to a crowded area. Any construction site fire will likely cause public panic and hurt the corporate image.

The major causes of construction site fires are electrical faults, hot working conditions, smoking, temporary lighting and lamp failures, lightning, and arson.

The risk of a fire breaking out on a construction site is always an important consideration. With all the exposed raw materials and tools in such close proximity to one another and not enough eyes to monitor large areas, there is a rising demand for construction site fire detectors.

The most common form of construction site fire detection is the smoke detector, which has a few limitations. Particles of combustion or smoke from a developing fire may not reach the sensing fire detector because the amount of “smoke” present may be insufficient to alarm smoke detectors. Barriers such as doors and walls may inhibit the flow and smoke particles may become “cold” and stratify, and may not reach the detectors smoke, or particles may be blown away from detectors by air outlets.

A better solution for AI visual smoke detection?

Several construction companies in Hong Kong contacted RoboticsCats in the last year, sharing their needs for a more effective construction site fire detection system. They asked if the visual-based AI wildfire detection could detect a construction site fire.

It was a very interesting question that excited RoboticsCats to find the answer.  The team did some lab tests to detect simulated construction site fire breakouts. The results were very encouraging. They trained their Machine Learning object detection algorithm using the videos provided by the construction companies. They conducted fire tests at several construction sites with the AI algorithm successfully detecting the fires!

With support from the construction companies, they developed a few pilots in Hong Kong, using high-quality video surveillance cameras to capture HD video images in both daytime and night-time. To move the cameras around easily to work in the rapid changing construction site environment, they use 5G/4G/LTE cellular routers to connect the cameras to the Internet. The AI-Cloud fire detection SaaS optimised detection algorithms for construction site fires. ReportFires app is the client software to receive construction site fire alerts.

The AI visual smoke detection of construction site fires is new and very promising. Both wildfire and construction site fires generate unnecessary carbon emissions and financial losses. The current pilots help customers and RoboticsCats to develop a better solution for the construction industry.

Check out the INSIGHT FD 3  Fire Detection System by RoboticsCats.