I&T Solution

Reference No.

S-0836

Solution Name

AI visual based human traffic detection

Solution Description

本方案採用邊緣運算(Edge Computing)設備運行人工智能視覺檢測技術,實時監測人流及場地狀況。系統能精準提供特定通道內上行與下行的雙向人流數量,以及區域內的總人數。方案亦可靈活擴充並整合其他視覺模型,以檢測各種違規或潛在危機狀況(如違規吸煙、非法闖入禁區等)。 邊緣運算設備具備低功耗、充足算力及極簡易部署的顯著優勢,特別適合應用於短期性大型活動的人流監測,以及在遠足高峰期對特定地點(如郊野徑、公共燒烤場、營地)進行分區監控。 在私隱保障方面,基於本產品的「隱私設計」(Privacy by Design)原則,攝取的影像僅由邊緣運算平台進行自動化分析,全程不經人手處理。最終的分析結果只會以熱像圖、圖表、數字數據或警報形式呈現,系統絕不會存檔或錄影原始影像,全面且嚴格地保障市民個人私隱。

This solution utilises edge computing devices powered by Artificial Intelligence (AI) visual detection technology to monitor human traffic and venue conditions in real time. The system accurately provides upstream and downstream directional flow counts, as well as the total headcount within a specific area or passage. It can also be flexibly integrated with additional visual models to detect various violations or potential hazardous situations (e.g., unauthorised smoking, trespassing into restricted zones).

Featuring low power consumption, sufficient computing power, and extremely easy deployment, our edge computing devices are highly suitable for crowd monitoring in short-term large-scale events and zoned surveillance at specific locations (such as hiking trails, public BBQ pits, and campsites) during peak seasons. Regarding privacy protection, adhering to the "Privacy by Design" principle of our product, captured images are solely analysed automatically by the edge computing platform without any human processing. The final analytical data is presented exclusively as heatmaps, graphs, numerical data, or alerts. No original images or videos will be archived or recorded, ensuring comprehensive and strict protection of public personal privacy.

Application Areas

City Management

Development

Health

Recreation and Culture

Technologies Used

Artificial Intelligence (AI)

Computer Vision

Data Analytics

Image Recognition

Predictive Analytics

Use Case

If any government department would like to conduct PoC trial or technology testing on the I&T solution, please contact Smart LAB.