Intelligent Video Monitoring for Real-Time Detection and Recognition of Elderly Falls on the Embedded Platform

With the ageing of the social population, in order to avoid elderly people living alone at home because of accidental falls and not timely treatment, this article put forward a computer vision based on the elderly fall detection and recognition system. Using the YOLOv3-tiny algorithm and DeepSORT al...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:2023 IEEE International Conference on Image Processing and Computer Applications (ICIPCA) s. 630 - 635
Hlavní autoři: Fan, Shicheng, Li, Mingde, Han, Chi
Médium: Konferenční příspěvek
Jazyk:angličtina
Vydáno: IEEE 11.08.2023
Témata:
Biological system modeling
Classification algorithms
Convolution
DeepSORT
Embedded System
FastPose
Hardware
Human Fall Detection
Older adults
Real-time systems
ST-GCN
Streaming media
YOLOv3-tiny
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:With the ageing of the social population, in order to avoid elderly people living alone at home because of accidental falls and not timely treatment, this article put forward a computer vision based on the elderly fall detection and recognition system. Using the YOLOv3-tiny algorithm and DeepSORT algorithm as the elderly detection and tracking stage, the system uses the FastPose algorithm and Spatio-Temporal Graph Convolution (ST-GCN) network to load the key points of the identified elderly human body and classify the elderly's actions. Then, Ascend MindX SDK is used to reason the model and deployed on the Atlas 200DK hardware platform so as to realize the purpose of real-time human motion detection. The comparison of the test results to those of other detection algorithms and the YOLOv3-tiny network model demonstrates that the detection algorithm has a good recognition effect.
DOI:10.1109/ICIPCA59209.2023.10257766