The Hong Kong Fire Services Department (FSD) is a government body specializing in public safety and disaster response. Tasked with protecting lives and reducing harm during emergencies, the department plays a vital role in ensuring community safety. As the nature of emergencies evolves, the department continually seeks innovative approaches to enhance its search and rescue capabilities.
PUBLIC SECTOR
Challenges
The FSD faces an array of challenges in its search and rescue missions, which often occur under extreme conditions such as mountainous terrain, dense forests, cliffs, and high temperatures. A significant hurdle was the accuracy compromise in predicting search zones, affected by the absence of a centralized platform for data aggregation and analysis.
Another obstacle was the limited availability of training datasets, which hampered the development of advanced AI deep learning models for more accurate image analysis. The lack of data restricted the department’s ability to improve its technology over time. These challenges highlighted the importance of adopting innovative solutions to strengthen the department's operational efficiency and optimize its search and rescue efforts.
Solutions
To address these challenges, LPS developed an AI-assisted drone image analysis system. The solution utilizes a proprietary AI neural network to process drone-captured images and provide real-time actionable insights. Built with Python, C++, and deep learning frameworks, it delivers advanced features such as thermographic image analysis, human object detection, color recognition, infrared detection, and real-time video streaming.
The system enables the identification of individuals in need of rescue even in dense or obscured environments. It also includes predictive route planning, using AI to calculate efficient search paths based on heatmaps, timing, and geographical data.
Additionally, the solution has centralized data management, consolidating information from multiple sources for faster and more accurate analysis, either on-site or remotely through the central command team.
DOWNLOAD PDF
