Name of Application Service

Nation's First AI Human Form Recognition System: "Loudspeaker Rescuer" Safeguards Public Safety

Name of Unit

Tourism Information Department, Chiayi City Government

Hyperlink

Name of the Collaborative Group

Fire Bureau, Chiayi City Government

Police Bureau, Chiayi City Government

Enterprise Business Group, Chunghwa Telecom Co., Ltd.

Taiwan Sing Po Security Co., Ltd.

Hyperlink

Brief of the Application or Service

To ensure the safety of the Lantan Reservoir and prevent unauthorized access to restricted water areas, the Chiayi City Government has implemented the "Water Area AI Human Form Recognition System" across six key hotspots along the Lantan Lake Circular Trail. Equipped with 19 surveillance cameras, the system actively detects human presence in restricted water zones. Once detected, a "Loudspeaker Rescuer" system broadcasts real-time verbal warnings to immediately deter individuals, effectively preventing potential accidents.

This project integrates water safety measures, geofencing, human activity recognition, edge computing technology, communication software, and on-site broadcasting equipment into a comprehensive system. The system is connected to a security company's control center for centralized monitoring and response. In case of emergencies, rescue requests can be promptly relayed to the Police and Fire Bureaus. This innovative solution has significantly reduced incidents of public encroachment into restricted water areas and is the first of its kind in the nation.

The Chiayi City Government demonstrated exceptional cross-sector integration and communication collaboration capabilities, successfully facilitating an interdisciplinary partnership to develop an innovative water safety solution. This groundbreaking initiative has earned the prestigious "2025 Smart City Innovative Application Award."

Although various agencies are partially involved in water safety operations, integration across these units has proven to be a challenge.

Illustration of challenges in this project

Brief of Innovation Initial and Achievement

Lantan Reservoir, spanning 70 hectares, serves as a vital recreational hub and a crucial water source for Chiayi City. Renowned for its scenic beauty, often referred to as "Lantan Reflecting the Moon", the reservoir has unfortunately been associated with frequent incidents of people falling into the water and various eerie, unfounded local legends. These factors have contributed to negative perceptions, impacting Chiayi City's tourism brand and overall urban image.

This project, based on statistical data from the Fire Bureau Duty Command Center, involved the installation of 19 human activity recognition systems and loudspeaker broadcast systems across six high-risk drowning zones around the Lantan Reservoir. These systems were also integrated with private security companies. With a modest investment of NT$2.1 million, since its launch on April 1, 2024, the average monthly number of people inadvertently entering the water has decreased significantly by 34.6 times, effectively reducing both the frequency of fire and rescue personnel dispatches and the likelihood of drowning incidents at the source.

This project demonstrates high scalability and replicability, inspiring adoption by the Taichung City Government, Tainan City Government, and Kaohsiung City Government, all of which are preparing to implement similar water safety prevention measures. In addition to enhancing public safety by saving lives, this initiative also boosts the potential market opportunities for Taiwan's image recognition industry, further driving innovation and development in this sector.

Challenges and Solutions for the Application Deployment

I.          Challenges Encountered:

(I)       Complex administrative management: At least five different land management agencies oversee areas surrounding the Lantan Reservoir.

(II)       Frequent false alarms: Infrared blocking equipment is highly susceptible to environmental factors, reducing accuracy.

(III)     Water range definition: Significant fluctuations in reservoir water levels make it challenging to define monitoring boundaries.

(IV)    Limited workforce: Public service personnel are insufficient to manage the vast water area effectively.

II.        Solutions Implemented:

(I)       Cross-agency coordination and integration: Established the primary management unit for the site as the competent authority; conducted on-site drills.

(II)     Human form recognition system: Adopted advanced technology after visiting the Smart City Summit & Expo, introducing mature and reliable image recognition systems.

(III)    Reference data: The system dynamically adjusts its monitoring range according to water level fluctuations, ensuring precise and effective surveillance.

(IV)   Outsourcing to security firms: A professional security service provider offers 24/7 monitoring and alert capabilities, significantly reducing the burden on public sector personnel.

Innovation Index

(I)       Nationwide first: Integrating geofencing, human form recognition, communication software, and loudspeaker systems, this innovative approach has set a benchmark, inspiring widespread adoption across various sectors.

(II)     Cross-sector collaboration: The project brings together diverse teams from tourism, police, and fire bureaus, as well as private security firms.

(III)    System architecture: Utilizing edge computing, wireless transmission, and cloud storage, the system is developed around an AI-based image recognition module for enhanced accuracy and functionality.

The project was featured as a cover story in the March issue of the Fire Safety Monthly

Kaohsiung City Government representatives and Chiayi City Government team in a group photo.

Tainan City Government representatives and the Chiayi City Government team in a group photo.

Future Plan

To further monitor public activity in and around the water areas, additional 360-degree panoramic cameras will be installed, complemented by the establishment of a robust fiber optic network to expand monitoring coverage and ensure sufficient bandwidth. Drawing on the experience gained from this project, similar systems will be implemented along bicycle paths to prevent unauthorized motor vehicles from entering. Beyond recognizing "human forms," the system's capabilities will also extend to identifying incidents involving unauthorized entry by vehicles such as cars, motorcycles, and bicycles.