Real-Time Human Group Detection and Clustering in Crowded Environments Using Enhanced Multi-Object Tracking
Group detection is a critical yet challenging task in video-based applications such as surveillance analysis, especially in crowded and dynamic environments where complex pedestrian interactions occur. Traditional trajectory-based methods often struggle with occlusions and overlapping behaviors, lea...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2024-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10759659/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1846126375664091136 |
|---|---|
| author | Hyunmin Lee Donggoo Kang Hasil Park Sangwoo Park Dasol Jeong Joonki Paik |
| author_facet | Hyunmin Lee Donggoo Kang Hasil Park Sangwoo Park Dasol Jeong Joonki Paik |
| author_sort | Hyunmin Lee |
| collection | DOAJ |
| description | Group detection is a critical yet challenging task in video-based applications such as surveillance analysis, especially in crowded and dynamic environments where complex pedestrian interactions occur. Traditional trajectory-based methods often struggle with occlusions and overlapping behaviors, leading to inaccurate group identification. To address these limitations, we propose a novel algorithm that integrates an optimized YOLOv8 model with DeepSORT tracking, enhancing both detection accuracy and real time performance. Our approach uniquely combines high-precision object detection with stable multi-object tracking, ensuring consistent identification of individuals and groups over time, even in high-density scenarios. Additionally, we introduce an innovative method of constructing an adjacency matrix by integrating Euclidean distances and bounding box diagonal ratios, which is transformed into a graph to intricately analyze and predict complex group dynamics in real time. Experimental results on real-world airport CCTV footage demonstrate that our method significantly outperforms existing approaches, achieving higher precision and recall rates. Furthermore, the algorithm operates efficiently on standard hardware, indicating strong practical feasibility for real-time applications in public spaces. While challenges such as misclassification due to incomplete data annotations and occlusions remain, our study showcases the potential of integrating spatial and temporal data to advance real-time group detection and tracking, aiming to improve crowd management systems in public spaces. |
| format | Article |
| id | doaj-art-b56c1a00fa864e1f842d03d991d7b08b |
| institution | Kabale University |
| issn | 2169-3536 |
| language | English |
| publishDate | 2024-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj-art-b56c1a00fa864e1f842d03d991d7b08b2024-12-13T00:00:39ZengIEEEIEEE Access2169-35362024-01-011218402818403910.1109/ACCESS.2024.350366110759659Real-Time Human Group Detection and Clustering in Crowded Environments Using Enhanced Multi-Object TrackingHyunmin Lee0Donggoo Kang1https://orcid.org/0000-0001-6959-1361Hasil Park2https://orcid.org/0000-0001-9882-6094Sangwoo Park3Dasol Jeong4Joonki Paik5https://orcid.org/0000-0002-8593-7155Department of Artificial Intelligence, Chung-Ang University, Seoul, South KoreaDepartment of Image, Chung-Ang University, Seoul, South KoreaDepartment of Image, Chung-Ang University, Seoul, South KoreaDepartment of Image, Chung-Ang University, Seoul, South KoreaDepartment of Image, Chung-Ang University, Seoul, South KoreaDepartment of Artificial Intelligence, Chung-Ang University, Seoul, South KoreaGroup detection is a critical yet challenging task in video-based applications such as surveillance analysis, especially in crowded and dynamic environments where complex pedestrian interactions occur. Traditional trajectory-based methods often struggle with occlusions and overlapping behaviors, leading to inaccurate group identification. To address these limitations, we propose a novel algorithm that integrates an optimized YOLOv8 model with DeepSORT tracking, enhancing both detection accuracy and real time performance. Our approach uniquely combines high-precision object detection with stable multi-object tracking, ensuring consistent identification of individuals and groups over time, even in high-density scenarios. Additionally, we introduce an innovative method of constructing an adjacency matrix by integrating Euclidean distances and bounding box diagonal ratios, which is transformed into a graph to intricately analyze and predict complex group dynamics in real time. Experimental results on real-world airport CCTV footage demonstrate that our method significantly outperforms existing approaches, achieving higher precision and recall rates. Furthermore, the algorithm operates efficiently on standard hardware, indicating strong practical feasibility for real-time applications in public spaces. While challenges such as misclassification due to incomplete data annotations and occlusions remain, our study showcases the potential of integrating spatial and temporal data to advance real-time group detection and tracking, aiming to improve crowd management systems in public spaces.https://ieeexplore.ieee.org/document/10759659/Multi-object trackingvisual surveillancegroup detection |
| spellingShingle | Hyunmin Lee Donggoo Kang Hasil Park Sangwoo Park Dasol Jeong Joonki Paik Real-Time Human Group Detection and Clustering in Crowded Environments Using Enhanced Multi-Object Tracking IEEE Access Multi-object tracking visual surveillance group detection |
| title | Real-Time Human Group Detection and Clustering in Crowded Environments Using Enhanced Multi-Object Tracking |
| title_full | Real-Time Human Group Detection and Clustering in Crowded Environments Using Enhanced Multi-Object Tracking |
| title_fullStr | Real-Time Human Group Detection and Clustering in Crowded Environments Using Enhanced Multi-Object Tracking |
| title_full_unstemmed | Real-Time Human Group Detection and Clustering in Crowded Environments Using Enhanced Multi-Object Tracking |
| title_short | Real-Time Human Group Detection and Clustering in Crowded Environments Using Enhanced Multi-Object Tracking |
| title_sort | real time human group detection and clustering in crowded environments using enhanced multi object tracking |
| topic | Multi-object tracking visual surveillance group detection |
| url | https://ieeexplore.ieee.org/document/10759659/ |
| work_keys_str_mv | AT hyunminlee realtimehumangroupdetectionandclusteringincrowdedenvironmentsusingenhancedmultiobjecttracking AT donggookang realtimehumangroupdetectionandclusteringincrowdedenvironmentsusingenhancedmultiobjecttracking AT hasilpark realtimehumangroupdetectionandclusteringincrowdedenvironmentsusingenhancedmultiobjecttracking AT sangwoopark realtimehumangroupdetectionandclusteringincrowdedenvironmentsusingenhancedmultiobjecttracking AT dasoljeong realtimehumangroupdetectionandclusteringincrowdedenvironmentsusingenhancedmultiobjecttracking AT joonkipaik realtimehumangroupdetectionandclusteringincrowdedenvironmentsusingenhancedmultiobjecttracking |