Closed-circuit Television Footage Research Articles

UNCANNY ACTIVITY DETECTION USING CCTV MONITORING

"Uncanny Activity Detection Using CCTV Monitoring" introduces an innovative approach to support public safety through the integration of Machine Learning (ML) in real-time monitoring. In the contemporary world, Closed-Circuit Television (CCTV) surveillance stands as a fundamental and highly effective security measure for various premises, including hospitals, malls, and universities. It serves as a widely recognized tool for preventing and detecting unwanted activities. However, envisioning an public space equipped with several CCTV cameras across multiple buildings presents a logistical challenge. The manual monitoring of events across this expansive network is practically impossible. Furthermore, searching for a specific event in recorded video footage, even after it has occurred, proves to be a time-consuming endeavor. This project addresses the need for an efficient solution to manage and analyze extensive CCTV footage in complex environments, optimizing security practices and response times. Keywords: Anomaly Detection, Deep Learning, Human Behavior Recognition.

CCTV Footage De-identification for Privacy Protection: A Comprehensive Survey

<p>Privacy preservation is a significant concern in our data-driven society, both in the social and political spheres. These concerns are amplified by the continuous emergence of technologies and services, expanding the boundaries of what is possible in our modern era. Technologies, such as audio-video surveillance, play an essential role in security and law enforcement but can potentially lead to privacy breaches. One of the approaches for privacy preservation in still images and video is de-identification which involves replacing or concealing personal identifiers to prevent unauthorized, unintended use or disclosure of personal information. This paper comprehensively reviews privacy and security issues in closed-circuit television (CCTV) data streams, representative state-of-the-art de-identification techniques for privacy protection. We also present a service scenario in which a MEC-enabled distributed system is used for CCTV footage de-identification and re-identification for individuals’ privacy protection. Finally, we highlight open research challenges and future directions for privacy protection in CCTV footage.</p> <p> </p>

Vision-based real-time process monitoring and problem feedback for productivity-oriented analysis in off-site construction

The widespread adoption of surveillance cameras in work environments has enabled the direct and non-intrusive detection of productivity-related issues in the field of construction. In this research, a process monitoring and problem feedback framework is developed based on closed-circuit television footage and computer vision analysis to achieve real-time visual control of the work process and facilitate data-driven decision-making in off-site construction. To enhance the automation of productivity-related problem recognition, a novel video analysis algorithm is developed to process the inputted video footage and provide feedback with respect to nine productivity issues. The z-score and Exponential Moving Average methods are employed to eliminate detection errors, and the spatial density analysis method is adopted to visually analyze spatial information. The observed performance of the proposed framework demonstrates that it can accurately acquire data from footage and provide process monitoring and problem feedback in real time.

Anomaly Detection in Heterogeneous Data Using Graph Convolutional Networks

Closed-circuit television (CCTV) surveillance systems are ubiquitous tools for ensuring public and private safety. However, manual monitoring of surveillance footage is arduous and time-consuming, often resulting in delayed identification and response to potential threats. In this paper, we present a novel real-time threat detection system tailored for CCTV surveillance, employing cutting-edge deep learning methodologies, particularly graph convolutional networks (GCNs), to detect and classify high-movement levels within video frames. By treating videos as segments and delineating between anomalous (potentially threatening) and normal (safe) segments, our system facilitates continuous real-time monitoring of surveillance footage, enabling the timely identification of potential threats such as abuse, burglaries, explosions, shootings, fighting, shoplifting, road accidents, arson, robbery, stealing, assault, and vandalism. To assess the efficacy of our system, we conducted extensive experiments on a large-scale dataset comprising CCTV footage, yielding promising results. Additionally, we introduce a novel, comprehensive dataset comprising 128 hours of real-world surveillance videos, encompassing various anomalies and normal activities. This dataset serves dual purposes: general anomaly detection, and the recognition of specific anomalous activities. Our experimental findings demonstrate the superior performance of our method in anomaly detection compared to state-of-the-art approaches. Furthermore, we provide an analysis of several recent deep learning baselines for anomalous activity recognition, revealing the challenging nature of our dataset and the potential it holds for future research endeavors. Through the integration of advanced deep learning techniques and comprehensive dataset creation, our proposed system stands to significantly enhance the efficiency and effectiveness of CCTV surveillance, facilitating faster response times and bolstering security for individuals in diverse environments.

Deep learning based multimodal urban air quality prediction and traffic analytics

Urban activities, particularly vehicle traffic, are contributing significantly to environmental pollution with detrimental effects on public health. The ability to anticipate air quality in advance is critical for public authorities and the general public to plan and manage these activities, which ultimately help in minimizing the adverse impact on the environment and public health effectively. Thanks to recent advancements in Artificial Intelligence and sensor technology, forecasting air quality is possible through the consideration of various environmental factors. This paper presents our novel solution for air quality prediction and its correlation with different environmental factors and urban activities, such as traffic density. To this aim, we propose a multi-modal framework by integrating real-time data from different environmental sensors and traffic density extracted from Closed Circuit Television footage. The framework effectively addresses data inconsistencies arising from sensor and camera malfunctions within a streaming dataset. The dataset exhibits real-world complexities, including abrupt camera or station activations/deactivations, noise interference, and outliers. The proposed system tackles the challenge of predicting air quality at locations having no sensors or experiencing sensor failures by training a joint model on the data obtained from nearby stations/sensors using a Particle Swarm Optimization (PSO)-based merit fusion of the sensor data. The proposed methodology is evaluated using various variants of the LSTM model including Bi-directional LSTM, CNN-LSTM, and Convolutions LSTM (ConvLSTM) obtaining an improvement of 48%, 67%, and 173% for short-term, medium-term, and long-term periods, respectively, over the ARIMA model.

Analysis and comparison of various deep learning models to implement suspicious activity recognition in CCTV surveillance

The paper aims to analyze and compare various deep learning (DL) algorithms in order to develop a Suspicious Activity Recognition (SAR) system for closed-circuit television (CCTV) surveillance. Automated systems for detecting and classifying suspicious activities are crucial as technology’s role in safety and security expands. This paper addresses these challenges by creating a robust SAR system using machine learning techniques. It analyzes and compares evaluation metrics such as Precision, Recall, F1 Score, and Accuracy using various deep learning methods (convolutional neural network (CNN), Long short-term memory (LSTM) – Visual Geometry Group 16 (VGG16), LSTM – ResNet50, LSTM – EfficientNetB0, LSTM – InceptionNetV3, LSTM – DenseNet121, and Long-term Recurrent Convolutional Network (LRCN)). The proposed system improves threat identification, vandalism deterrence, fight prevention, and video surveillance. It aids emergency response by accurately classifying suspicious activities from CCTV footage, reducing reliance on human security personnel and addressing limitations in manual monitoring. The objectives of the paper include analyzing existing works, extracting features from CCTV videos, training robust deep learning models, evaluating algorithms, and improving accuracy. The conclusion highlights the superior performance of the LSTM-DenseNet121 algorithm, achieving an overall accuracy of 91.17% in detecting suspicious activities. This enhances security monitoring capabilities and reduces response time. Limitations of the system include subjectivity, contextual understanding, occlusion, false alarms, and privacy concerns. Future improvements involve real-time object tracking, collaboration with law enforcement agencies, and performance optimization. Ongoing research is necessary to overcome limitations and enhance the effectiveness of CCTV surveillance.

Deep Learning-Based Video Compression for Surveillance Footage

The utilization of closed-circuit television (CCTV) monitoring plays a pivotal role in the realm of video processing, providing an effective means for vigilant surveillance. However, a significant challenge associated with this practice is the substantial demand it places on storage space. Traditionally, surveillance footage is stored on hard disk drives, and due to limited storage capacities, it often necessitates periodic deletion. To tackle this issue, we have introduced an innovative method for compressing CCTV video, named “Detection-Based video Compression” (DBVC). Our DBVC model is a two-step process. In the first step, we employ advanced neural network approaches such as Mask-RCNN and YOLOv4 to determine active and idle frames within the surveillance video. These cutting-edge techniques enable precise identification of objects and events of interest in the video feed. In the second step, we construct a new video composed solely of the active frames, eliminating redundant or uneventful segments. After conducting a comprehensive analysis of the experimental results, it is evident that Mask-RCNN stands out with an impressive detection accuracy of 98% on the COCO dataset, making it a robust choice for identifying objects and events in the surveillance footage. Consequently, we chose to leverage the output generated by Mask-RCNN and YOLOv4 for subsequent processing stages. Our DBVC approach is a breakthrough in video compression technology, significantly reducing the storage space required for CCTV footage. In fact, it achieves an average compression ratio of up to 85% when using YOLOv4, surpassing the capabilities of existing state-of-the-art compression methods. This innovation not only optimizes storage efficiency but also maintains a high level of surveillance data integrity, making it a valuable advancement in the field of CCTV video processing and storage management.

Early fire detection technique for human being using deep learning algorithm

Fire and smoke detection in today’s world is a must, especially in clustered areas where a quick response can prevent significant damages and save lives. Early detection plays a significant role in preventing the fire from spreading by alerting the emergency response personnel. It may not be possible to install traditional fire and smoke detectors everywhere. As a result, incorporating fire and smoke detection into existing closed circuit television (CCTV) systems in various places can provide a warning to the appropriate authorities, allowing for quick action to prevent the fire from spreading. This work aims in developing an early fire and smoke prediction model with CCTV footage images and video frames. The images and videos are collected from multiple datasets available online. A convolutional neural network (CNN) model is developed for early detection and prevention of the spreading of fire and compares it with transfer learning models ResNet50 and VGG19. The model obtain an accuracy of around 94% using CNN model, 95% using VGG19 and 98% using ResNet 50. A model with high accuracy can replace traditional fire detection systems which can be both cost-effective and easy to implement to existing surveillance cameras.

Fusion of CCTV Video and Spatial Information for Automated Crowd Congestion Monitoring in Public Urban Spaces

Crowd congestion is one of the main causes of modern public safety issues such as stampedes. Conventional crowd congestion monitoring using closed-circuit television (CCTV) video surveillance relies on manual observation, which is tedious and often error-prone in public urban spaces where crowds are dense, and occlusions are prominent. With the aim of managing crowded spaces safely, this study proposes a framework that combines spatial and temporal information to automatically map the trajectories of individual occupants, as well as to assist in real-time congestion monitoring and prediction. Through exploiting both features from CCTV footage and spatial information of the public space, the framework fuses raw CCTV video and floor plan information to create visual aids for crowd monitoring, as well as a sequence of crowd mobility graphs (CMGraphs) to store spatiotemporal features. This framework uses deep learning-based computer vision models, geometric transformations, and Kalman filter-based tracking algorithms to automate the retrieval of crowd congestion data, specifically the spatiotemporal distribution of individuals and the overall crowd flow. The resulting collective crowd movement data is then stored in the CMGraphs, which are designed to facilitate congestion forecasting at key exit/entry regions. We demonstrate our framework on two video data, one public from a train station dataset and the other recorded at a stadium following a crowded football game. Using both qualitative and quantitative insights from the experiments, we demonstrate that the suggested framework can be useful to help assist urban planners and infrastructure operators with the management of congestion hazards.

Video analytics for optimizing bank services

The use of Closed-Circuit Television (CCTV) footage for surveillance, demographic monitoring, and behavior analysis has increased dramatically in the recent years. One of the key challenges in analyzing such recordings is to count people and identify emotions efficiently and accurately. This paper proposed a system for counting people and identifying facial expressions of emotion in CCTV footage. It used Deep Neural Networks (DNN) trained on massive annotation datasets to recognize people in videos, extract facial images, and identify emotions. It evaluated the performance of the method on publicly available datasets and compares it with state-of-the-art techniques. The findings demonstrate the ability of the method to accurately count people and identify their emotions.

공공영역에서의 개인정보 수집과 활용을 위한 제도적 개선방안

In recent years, artificial intelligence technology, which enables new technologies to be deployed, has expanded the scope and methods of data collection and utilization of government agencies. Data analysis and AI technology are deployed to enhance the intelligent and accurate use of Closed-circuit television (CCTV) surveillance cameras for crime prevention, security, and public safety purposes. The use of facial recognition technology enables an individual to be identified by processing images captured on CCTV, the suspects to be tracked on the CCTV footage, and crime scenes to be monitored in real time. Amidst this expansion, there is an increasing need for discussions to protect privacy. The need for privacy control is emphasized as data collection, retention, and utilization capabilities resulting in losing control over individual’s own information. The value and importance of personal data have been changed from the previous, and government surveillance issues arises as vast quantities of personal data being stored and easily shared. The need for government data access for various purposes and storage of data on a large scale may lead to the privacy right and the infringement of informational self-determination. In a situation where the use of CCTV information is allowed and gradually expanding, it is difficult to evaluate that legal grounds for controlling the actions of government agencies or guaranteeing the rights of individual subjects are sufficient. This paper aims to examine the current status of data collection by government institutions, especially the institutional basis for the use of CCTV and video information collection by government agencies, and problems with the system regarding personal data protection.

HUMAN ABNORMAL ACTIVITY RECOGNITION

This paper brings forward one amongst the foremost significant applications of human suspicious activity recognition that is termed as anomaly detection. A key concern of any society today is providing safety to an individual. The main reason behind this concern is due to the constantly increasing activities causing threats, starting from deliberate ferocity to an injury caused through an accident. Simple installation of a traditional closed circuit television(CCTV) is not sufficient as it requires a person to continuously stay alert and monitor the cameras, which is quite inefficient. This call for the requirement to develop an security system which is fully automated system that recognizes anomalous activities in real time and brings instant help to the victims. Hence we proposed a system which will examine and detect the suspicious human action from real-time CCTV footage with help of machine learning techniques and generates the alert if the abnormal activity is occurred. The method is implemented on the dataset containing both normal and anomaly activity and experiment has shown better results.

Counting people inside a region-of-interest in CCTV footage with deep learning.

In recent years, the performance of people-counting models has been dramatically increased that they can be implemented in practical cases. However, the current models can only count all of the people captured in the inputted closed circuit television (CCTV) footage. Oftentimes, we only want to count people in a specific Region-of-Interest (RoI) in the footage. Unfortunately, simple approaches such as covering the area outside of the RoI are not applicable without degrading the performance of the models. Therefore, we developed a novel learning strategy that enables a deep-learning-based people counting model to count people only in a certain RoI. In the proposed method, the people counting model has two heads that are attached on top of a crowd counting backbone network. These two heads respectively learn to count people inside the RoI and negate the people count outside the RoI. We named this proposed method Gap Regularizer and tested it on ResNet-50, ResNet-101, CSRNet, and SFCN. The experiment results showed that Gap Regularizer can reduce the mean absolute error (MAE), root mean square error (RMSE), and grid average mean error (GAME) of ResNet-50, which is the smallest CNN model, with the highest reduction of 45.2%, 41.25%, and 46.43%, respectively. On shallow models such as the CSRNet, the regularizer can also drastically increase the SSIM by up to 248.65% in addition to reducing the MAE, RMSE, and GAME. The Gap Regularizer can also improve the performance of SFCN which is a deep CNN model with back-end features by up to 17.22% and 10.54% compared to its standard version. Moreover, the impacts of the Gap Regularizer on these two models are also generally statistically significant (P-value < 0.05) on the MOT17-09, MOT20-02, and RHC datasets. However, it has a limitation in which it is unable to make significant impacts on deep models without back-end features such as the ResNet-101.

ACF: An Armed CCTV Footage Dataset for Enhancing Weapon Detection.

Thailand, like other countries worldwide, has experienced instability in recent years. If current trends continue, the number of crimes endangering people or property will expand. Closed-circuit television (CCTV) technology is now commonly utilized for surveillance and monitoring to ensure people’s safety. A weapon detection system can help police officers with limited staff minimize their workload through on-screen surveillance. Since CCTV footage captures the entire incident scenario, weapon detection becomes challenging due to the small weapon objects in the footage. Due to public datasets providing inadequate information on our interested scope of CCTV image’s weapon detection, an Armed CCTV Footage (ACF) dataset, the self-collected mockup CCTV footage of pedestrians armed with pistols and knives, was collected for different scenarios. This study aimed to present an image tilling-based deep learning for small weapon object detection. The experiments were conducted on a public benchmark dataset (Mock Attack) to evaluate the detection performance. The proposed tilling approach achieved a significantly better mAP of 10.22 times. The image tiling approach was used to train different object detection models to analyze the improvement. On SSD MobileNet V2, the tiling ACF Dataset achieved an mAP of 0.758 on the pistol and knife evaluation. The proposed method for enhancing small weapon detection by using the tiling approach with our ACF Dataset can significantly enhance the performance of weapon detection.

Does wearing a surgical mask influence face touching by healthcare workers? A retrospective cohort study.

At the start of the COVID-19 pandemic, healthcare workers (HCW) in our ED were advised against and actively discouraged from wearing masks when not seeing respiratory patients, as mask wearing was thought to increase the risk of droplet transmission by face touching. The primary objective of the present study was to determine whether HCW using face masks were more or less likely to touch their faces than those not wearing masks. We analysed six randomly selected hours of closed circuit television footage from our staff base. Face touches were recorded electronically by trained researchers. Generalised linear mixed models were used to compare the frequency and duration of face touches with and without face masks, controlling for individual clusters, adjusting for time of footage, duration on screen and staff role. Data were collected from 187 HCW. Masks were worn in 231 (36%) of 642 screen sessions. Wearing a mask did not significantly change the odds of face touching (odds ratio 0.55, 95% confidence interval [CI] 0.30-1.01, P=0.055) or duration of face touch (mean difference -1.45 s, 95% CI -8.84, 5.99, P=0.71). For mucosal areas, a significant reduction in the odds of face touching was observed for mask wearers (odds ratio 0.21, 95% CI 0.11-0.43, P < 0.001) and on the frequency of mucosal touches (rate ratio 0.45, 95% CI 0.29-0.69, P < 0.001). Mask wearing did not change face touching or the duration of face touches. However, significantly fewer mucosal touches were observed when wearing a mask, which may help to reduce nosocomial droplet transmission of viruses.

Early Response to COVID-19

Early Response to COVID-19

Setting foot in private spaces: extending the hepatitis C cascade of care to automatic needle/syringe dispensing machines, a mixed methods study

BackgroundGlobal commitment to achieving hepatitis C virus (HCV) elimination has enhanced efforts in improving access to direct-acting antiviral (DAA) treatments for people who inject drugs (PWID). Scale-up of efforts to engage hard-to-reach groups of PWID in HCV testing and treatment is crucial to success. Automatic needle/syringe dispensing machines (ADMs) have been used internationally to distribute sterile injecting equipment. ADMs are a unique harm reduction service, affording maximum anonymity to service users. This paper explores the feasibility and acceptability of extending the HCV cascade of care to sites where ADMs are located.MethodsThe ADM users into Treatment (ADMiT) study was undertaken in a metropolitan region in Sydney, Australia. This mixed methods study involved analysis of closed-circuit television footage, ethnographic methods (fieldwork observation and in-depth interviews) and structured surveys. Researchers and peers conducted fieldwork and data collection over 10 weeks at one ADM site, including offering access to HCV testing and treatment.ResultsFindings from 10 weeks of fieldwork observations, 70 survey participants and 15 interviews highlighted that there is scope for engaging with this population at the time they use the ADM, and enhanced linkage to HCV testing and treatment may be warranted. Most survey participants reported prior HCV testing, 61% in the last 12 months and 38% had received HCV treatment. However, fieldwork revealed that most people observed using the ADM were not willing to engage with the researchers. Field work data and interviews suggested that extending the HCV cascade of care to ADMs may encroach on what is a private space for many PWID, utilized specifically to avoid engagement.DiscussionEnhanced linkage to HCV testing and treatment for people who use ADMs may be warranted. However, data suggested that extending the HCV cascade of care to ADMs may encroach on what is a private space for many PWID, utilized specifically to avoid engagement. The current study raises important public health questions about the need to ensure interventions reflect the needs of affected communities, including their right to remain anonymous.

Particulate Matter Estimation from Public Weather Data and Closed-Circuit Television Images

This article proposes a method of estimating the concentrations of particulate matter (PM2.5 and PM10) using public data, including road-traffic closed-circuit television (CCTV) images, Smart Seoul City data sensor environment information, and Korea Meteorological Administrationdata. The region-of-interest images and full scenes derived from CCTV footage were used as the basis for the deep learning model, which combines a convolutional neural network and long short-term memory, to establish the particulate matter (PM) concentration prediction methodology. In the experiment, the prediction accuracies corresponding to various types of mean values were calculated by training the model with various mean measurement values for the surface PM2.5 and PM10 concentrations, as well as the corresponding CCTV images and weather data at different time points. In the experiments performed under relatively stable PM concentrations, R2 generally exceeded 0.9 and tended to increase with an increasing range of mean concentration values. In particular, in sections with rapid changes in the PM concentration within an hourly interval, higher R2 values were obtained by the model trained with the average PM concentrations of the time series before and after image capture, outperforming the method that used prior mean observation values and better reflecting the current PM concentration.

Efficacy Determination of Various Base Networks in Single Shot Detector for Automatic Mask Localisation in a Post COVID Setup

ABSTRACT The COVID-19 pandemic is one of the rarest events of global crises where a viral pathogen infiltrates every part of the world, leaving every country face an inevitable threat of having to lock down major cities and economic hubs and put firm restrictions on citizens thus slowing down the economy. The risk of removal of lockdowns is the emergence of new waves of a pandemic causing a surge in new cases. These facts necessitate the containment of the virus when the lockdowns end. Wearing masks in crowded places can help restrict the spread of the virus through minuscule droplets in the air. Through the automatic detection, enumeration, and localisation of masks from closed-circuit television footage, it is possible to keep violations of post-COVID regulations in check. In this paper, we leverage the Single-Shot Detection (SSD) framework through different base convolutional neural networks (CNNs) namely VGG16, VGG19, ResNet50, DenseNet121, MobileNetV2, and Xception to compare performance metrics attained by the different variations of the SSD and determine the efficacies for the best base network model for automatic mask detection in a post COVID world. We find that Xception performs best among all the other models in terms of mean average precision.

Successful control of COVID-19 outbreak through tracing, testing, and isolation: Lessons learned from the outbreak control efforts made in a metropolitan city of South Korea

The first surge of coronavirus disease 2019 (COVID-19) cases began on June 27, 2020 in Gwangju metropolitan city, located in the southwestern part of South Korea, with a population of 1,501,000. Local governments and the Korean Center for Disease Control and Prevention immediately started an epidemiologic investigation and traced the contacts of patients using a wide variety of data sources, including location data from mobile devices, credit card transaction, and closed-circuit television footage. Until July 16, 2020, 138 community transmission cases and 10 infection clusters were identified across the city. Through contact tracing, epidemiologic relatedness was found in 136 (98.6%) of 138 cases. Our investigation showed how the extensive and meticulous contact tracing suppressed COVID-19 outbreak in a populated city.