Automated Physical Distance Estimation and Crowd Monitoring Through Surveillance Video.

The contagious Corona Virus (COVID-19) transmission can be reduced by following and maintaining physical distancing (also known as COVID-19 social distance). The World Health Organisation (WHO) recommends preventing COVID-19 from spreading in public areas. On the other hand, people may not be maintaining the required 2-m physical distance as a mandated safety precaution in shopping malls and public places. The spread of the fatal disease may be slowed by an active monitoring system suitable for identifying distances between people and alerting them. This paper introduced a deep learning-based system for automatically detecting physical distance using video from security cameras. The proposed system introduced the TH-YOLOv5 for object detection and classification and Deepsort for tracking the detected people using bounding boxes from the video. TH-YOLOv5 included another prediction head to identify objects of varying sizes. The original prediction heads are then replaced with Transformer Heads (TH) to investigate the prediction capability of the self-attention mechanism. Then, we include the convolutional block attention model (CBAM) to identify attention areas in settings with dense objects. Pairwise L2 vectorized normalization was utilized to generate a three-dimensional feature space for tracking physical distances and the violation index, determining the number of individuals who follow the distance rules. We use the MS COCO and HumanCrowd, CityPersons, and Oxford Town Centre (OTC) data sets for training and testing. Experimental results demonstrate that the proposed system obtained a weighted mAP score of 89.5% and an FPS score of 29; both are computationally comparable.

A Deep-Learning Based Automated COVID-19 Physical Distance Measurement System Using Surveillance Video

The contagious Corona Virus (COVID-19) transmission can be reduced by following and maintaining physical distancing (also known as COVID-19 social distance). The World Health Organisation (WHO) recommends it to prevent COVID-19 from spreading in public areas. On the other hand, people may not be maintaining the required 2-m physical distance as a mandated safety precaution in shopping malls and public places. The spread of the fatal disease may be slowed by an active monitoring system suitable for identifying distances between people and alerting them. This paper introduced a deep learning-based system for automatically detecting physical distance using video from security cameras. The proposed system employed the fine-tuning YOLO v4 for object detection and classification and Deepsort for tracking the detected people using bounding boxes from the video. Pairwise L2 vectorized normalization was utilized to generate a three-dimensional feature space for tracking physical distances and the violation index, determining the number of individuals who follow the distance rules. For training and testing, we use the MS COCO and Oxford Town Centre (OTC) datasets. We compared the proposed system to two well-known object detection models, YOLO v3 and Faster RCNN. Our method obtained a weighted mAP score of 87.8% and an FPS score of 28;both are computationally comparable. © 2022, Springer Nature Switzerland AG.

Close-Contact Detection Using a Single Camera for Sports Considering Occlusion

The Coronavirus disease 2019 (COVID-19) is still prevalent in the world. Exercise is important to maintain our health while dealing with infectious diseases. Social distancing is more important during exercise because we may not be able to wear masks to avoid breathing problems, heatstroke, etc. To maintain social distancing during exercise, we develop a close-contact detection system using a single camera especially for sports in schools and gyms. We rely on a single camera because of the deployment cost. The system recognizes people from a video and estimates the interpersonal distance for close-contact detection. The challenge is the occlusion of people, which leads to false negatives in close-contact detection. To solve the problem, we leverage the observation that most false negatives in human detection are caused by occlusion owing to other people. This is because there are few obstacles in sports facilities. Based on the above observation, we assume that a person still exists near the last detected position even when s/he disappears in the proximity of other people. For evaluation, we recorded 834 videos that were 112 min long in total including various scenarios with 2724 close-contacts. The results show that the F1-score of close-contact detection and tracking are 83.6% and 67.3%, respectively. We also confirmed that the start and end time errors are within 1 s for more than 80% of the close-contacts. © 2013 IEEE.