A research team at the Sports Artificial Intelligence Laboratory in the Department of Electrical and Electronic Engineering, at the University of Hong Kong, has developed a highly promising innovation called the “Artificial Intelligence MGF Network for Anomalies Detection” technology.
The Sports AI Laboratory has developed the Glance and Focus AI Anomalies Detection technology, which surpasses current artificial intelligence detection methods limited to specific targets in static images. This innovative technology enables real-time analysis of human posture and movements in videos, allowing for precise and effective detection of abnormal situations.
It can accurately identify events such as falls, fainting and drowning, as well as instances of violence like fighting and abuse. By providing timely alerts, this technology aids in facilitating rescue efforts and preventing accidents from occurring.
The research team uses artificial intelligence and deep learning to develop an anomaly detection algorithm to detect skeleton joint points for estimations of movements and poses. The system can identify in real-time from about seven to eight moving frames in a video, i.e., about one-quarter of second, possible abnormal situations and raise alarm.
Using artificial intelligence and deep learning techniques, the research team has devised an anomaly detection algorithm for detecting skeleton joint points in order to estimate movements and poses. By employing this algorithm, the system can effectively identify potential abnormal situations in real-time video streams, analysing approximately seven to eight moving frames, equivalent to around a quarter of a second. Upon detecting such anomalies, the system promptly raises an alarm, enabling swift intervention and response.
Furthermore, the system demonstrates comparable accuracy and effectiveness when applied to thermal images. It can accurately detect body movements in thermal imagery, without relying on additional visual details. This aspect ensures the protection of personal privacy while effectively detecting anomalies. By leveraging thermal imaging, the system maintains its ability to identify abnormal situations while preserving the privacy of individuals being monitored.
The research team has forged partnerships with pertinent organisations to facilitate the application of their new technology. These collaborations extend to various sectors, including children and elderly care facilities, as well as swimming pools for drowning alerts.
Integrating the AI-based anomaly detection system into these settings enables enhanced safety measures and timely response capabilities. This technology has the potential to provide valuable support and assistance in ensuring the well-being and security of individuals in these environments.
With the backing of the Smart Traffic Fund, the research team is actively investigating the viability of employing thermal images to analyse pedestrian movements and postures at traffic light junctions. This innovative approach aims to enable smart extensions of crossing time for individuals who require additional assistance, such as the elderly, children, or people using wheelchairs, allowing them sufficient time to safely navigate the road.
Additionally, the team has plans to extend the application of this technology to bus terminals, where it can be utilised to provide warnings and alerts regarding potential hazardous situations involving pedestrians and other road users. By implementing this technology in these contexts, the team envisions enhancing pedestrian safety and improving overall traffic management.
The Director of the Sports AI Laboratory emphasised that this cutting-edge technology has the potential to save lives. The research team’s primary objective during its development was to address emergency situations and urgent applications in daily life, aiming to create a safer and more convenient living environment for the general public.
