Cerberus: Privacy-Preserving Crowd Counting and Localisation

University of Cambridge 2024

Brief Description

Cerberus employs face detection on edge devices for privacy-preserving crowd counting and localisation, designed for real-time applications in urban informatics and building management without compromising privacy.

Leveraging high-resolution cameras and optional hardware accelerators, this system integrates with digital twins for advanced data visualization and smart building applications.

More information

The Cerberus project introduces a method for real-time crowd monitoring and privacy preservation using edge computing devices with face detection technology. It is designed for applications in urban informatics and smart building management. The following sections provide an overview of the system architecture, performance comparison of machine learning models, and analysis of occupancy patterns, emphasizing privacy considerations.

Cerberus Poster
Best poster award, Cambridge Sensors Day 2023.

Presented at EdgeSys 2024 (EuroSys '24).

System Overview

Cerberus Device Configuration

The Cerberus device setup includes a Raspberry Pi 4 and a high-resolution Pi Camera, designed to leverage the compact yet powerful hardware for edge-based computations. This configuration is crucial for performing real-time face detection and occupancy analysis, allowing for efficient, local data processing without relying on cloud services. The optional integration of a hardware accelerator further boosts the system's ability to handle complex machine learning models for face detection, ensuring swift and accurate data analysis. This setup exemplifies the project's commitment to combining high-performance computing with privacy-preserving technologies in smart building applications.

The privacy preserving cameras allow to track human occupancy in the lecture theatre in real-time. The vide below shows a lecture taking place:

Cerberus Device Configuration

Sensor Placement and Coverage

Sensor Placement

Strategic placement of sensors is a foundational element of the Cerberus system, designed to maximize coverage and enhance the accuracy of occupancy detection and localization. By carefully positioning cameras and sensors throughout the monitored area, the system ensures that no significant blind spots exist, thereby improving the reliability of the data collected. This methodical approach to sensor deployment is critical for capturing a complete picture of space utilization, allowing for precise monitoring of crowd dynamics and individual movements within a given environment. The effectiveness of this strategy is evident in the system's ability to provide granular, actionable insights, essential for optimizing space management and ensuring safety in real-time.

System Architecture

Cerberus System Architecture

The integration of edge devices and high-resolution cameras forms the backbone of the Cerberus system, enabling local data processing directly on the devices. This architecture is key to maintaining the privacy of individuals within monitored spaces, as it eliminates the need for image storage or transmission over networks. By processing data locally, the system minimizes potential privacy risks associated with cloud computing and external data handling. This approach not only ensures the real-time performance of the system but also aligns with strict privacy regulations by design, making it an ideal solution for sensitive environments.

Model Performance

Model Comparison

The system employs various machine learning models for face detection, each with its unique trade-offs between inference speed and accuracy. This comparative analysis allows for a tailored approach to selecting the optimal model based on the specific requirements of different operational environments. High-accuracy models may be preferred in scenarios where precise detection is critical, albeit at the cost of slower processing times. Conversely, faster models that offer real-time performance may be chosen for environments where speed is paramount. This flexibility ensures that the Cerberus system can be effectively adapted to a wide range of settings, from densely populated public spaces to more controlled access areas, maintaining a balance between operational efficiency and the quality of surveillance.

User Interface and Data Visualization

Cerberus UI

The system features a sophisticated user interface designed for intuitive visualization of occupancy and crowd data. This interface serves as a critical tool for administrators and facility managers, offering an immediate, comprehensive view of space utilization and occupant distribution. Through dynamic charts, graphs, and heatmaps, users can interact with real-time data, gaining insights into patterns of movement, peak occupancy times, and areas of congestion. This level of data interpretation and interaction aids in effective space management, enabling data-driven decisions for building layout adjustments, emergency response planning, and enhancing overall occupant comfort and safety.

Occupancy Patterns

KDE Plot

Kernel Density Estimation (KDE) plots are employed to reveal intricate occupancy and movement patterns within the monitored environment, playing a crucial role in the analysis and optimization of space utilization. These plots provide a visual representation of how spaces are used over time, highlighting areas of high and low occupancy and identifying patterns of movement that may not be apparent from raw data alone. By analyzing these patterns, facility managers can make informed decisions about space planning, layout modifications, and occupancy limits to enhance efficiency and safety. The use of KDE plots underscores the system's advanced analytical capabilities, enabling a deeper understanding of how spaces function and how they can be improved.

Results and Future Directions

Over its year-long deployment, Cerberus has proven to be highly effective in privacy-preserving crowd monitoring, with potential applications extending to various public and private sectors for improved space management and emergency planning.

Future enhancements will focus on exploring more advanced machine learning models for face detection, improving system scalability, and further enhancing privacy protections through advanced encryption methodologies.