2021
DOI: 10.1080/0952813x.2021.1938696
|View full text |Cite
|
Sign up to set email alerts
|

A human fall detection framework based on multi-camera fusion

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
2

Citation Types

0
4
0

Year Published

2021
2021
2025
2025

Publication Types

Select...
4
2
1

Relationship

0
7

Authors

Journals

citations
Cited by 10 publications
(4 citation statements)
references
References 42 publications
0
4
0
Order By: Relevance
“…For example, several solutions have recently been proposed in the area of fall detection for remote monitoring of fragile patients. In [ 23 ], multi-camera fusion is performed by combining models trained on single cameras together into a global ensemble model at the decision-making level, providing higher accuracy with respect to local single-camera models and avoiding computationally expensive cameras calibration. The dual-stream fused neural network method, proposed in [ 24 ], first trains two deep neural networks to detect falls by using two single cameras and then merges the results through a weighted fusion of prediction scores.…”
Section: Related Workmentioning
confidence: 99%
See 1 more Smart Citation
“…For example, several solutions have recently been proposed in the area of fall detection for remote monitoring of fragile patients. In [ 23 ], multi-camera fusion is performed by combining models trained on single cameras together into a global ensemble model at the decision-making level, providing higher accuracy with respect to local single-camera models and avoiding computationally expensive cameras calibration. The dual-stream fused neural network method, proposed in [ 24 ], first trains two deep neural networks to detect falls by using two single cameras and then merges the results through a weighted fusion of prediction scores.…”
Section: Related Workmentioning
confidence: 99%
“…In particular, our solution exploits a transfer learning approach, which consists in training the object detection model on a single camera, in updating it through an additional training by feeding the other camera’s images, and then by fusing the single detection signals to generate alerts at the decision level. This speeds up the overall training time and saves computational resources, with respect to other existing decision-making level camera fusion approaches, such as [ 22 , 23 ].…”
Section: Related Workmentioning
confidence: 99%
“…An effective video surveillance system must be robust to different types of environments and different climate conditions and must provide useful information on the scene regardless of the area to be monitored (La & Ananth, 2022). Moreover, surveillance systems based on RADAR (maritime or aerial), not only are very expensive but they also suffer from many limitations, namely dead zones, their incapability of silhouette recognition and poor pedestrian detection (Ezatzadeh et al, 2021). Thus, a potential solution to this problem is the use of several multispectral imaging devices; however, the aggregation of several image modalities results in a significant increase in the amount of data.…”
Section: Introductionmentioning
confidence: 99%
“…Human detection is a computer task that has been in development for at least two decades now [1]. This technology has had various applications in search and rescue [2][3][4][5][6][7][8], law enforcement [9][10][11], pedestrian detection for traffic management and automated driver assistance [12][13][14][15][16][17][18][19], fall detection [20][21][22][23][24][25][26], and many other functions, including the decision-making steps that ensue [27]. And oftentimes, the technology is deployed through unmanned aerial vehicles (UAVs) [4,[6][7][8][28][29][30][31] given their flexibility, longer range of tracking, and ability to acquire images and videos in situations where acquisition is infeasible for cameras at the ground level [27,[32][33][34].…”
Section: Introductionmentioning
confidence: 99%