Revolutionizing Retail with AI-Driven Surveillance
About the Client
Company’s Request
Technology Set
|
|
|
|
Rockchip RK3399Pro
|
A development board with a high-performance NPU for AI tasks.
Chosen for its edge computing power, enabling real-time video analysis and decision-making for surveillance.
|
|
TensorFlow
|
An AI framework for developing and training machine learning models. Optimized AI models for the RK3399Pro NPU, enhancing efficiency and performance in surveillance tasks.
|
|
AWS IoT Core
|
A cloud service for device connectivity and management. Enabled centralized management of surveillance devices, facilitating updates and security.
|
|
AWS S3 and Lambda
|
S3 provide scalable storage; Lambda offers serverless computing for data processing. Stored large volumes of video data and processed information for advanced analytics.
|
|
Python
|
A programming language for software development. Used for developing the software components of our system.
|
|
OpenCV
|
A library for image processing and computer vision. Processed video feeds from cameras, key for real-time image analysis and surveillance operations.
|
|
PyTorch
|
An AI framework for deep learning model development and training. Employed alongside TensorFlow for training AI models, offering alternatives for optimization and efficiency.
|
|
Docker
|
A platform for developing, shipping, and running applications in containers. Used as a consistent environment for our application for deployment and scaling across devices.
|
Concept
The solution we developed is a modular, scalable system designed for flexibility and ease of expansion. Each module in this system is built around a Rockchip RK3399Pro board that has enough computation capabilities to support up to 10 surveillance cameras in real-time.
This setup allows for rapid integration with the cloud management system, ensuring that extending coverage or enhancing surveillance capabilities can be achieved with minimal effort and disruption.
When the client needs to expand their surveillance area, they can simply deploy an additional module. Each module can be quickly connected to the existing cloud management infrastructure, providing integration and centralized control over the expanded network of cameras. This modular design not only simplifies the process of scaling up but also provides the flexibility to adjust the number of cameras based on the specific requirements of each site.
The solution we developed was primarily designed to harness the edge computing capabilities of the Rockchip RK3399Pro development board, allowing all main computational tasks, especially those related to real-time surveillance analysis, to be executed on-site.
With this approach, we aimed to maximize the responsiveness and reliability of the security surveillance system by leveraging the board’s processing power to analyze video streams instantly, make decisions, and trigger alerts without delay. The key advantages of this setup included enhanced privacy, reduced bandwidth consumption, and operational continuity even in the event of network disruptions.
Cloud Integration
However, we strategically decided to incorporate cloud services into the solution for several reasons, despite the emphasis on edge computing:
Data Storage and Management: The cloud component was integrated to handle the extensive data storage needs that exceed the local capacity of the RK3399Pro. Surveillance systems generate vast amounts of video data; storing this data on the cloud ensures scalability and easy access for review and analysis without overwhelming the edge devices.
Device Management and Configuration: Utilizing AWS IoT Core allowed us to streamline the management of multiple surveillance devices deployed across different locations. This cloud service enables centralized control, firmware updates, and security patches, ensuring all devices operate efficiently to maintain the integrity of a large-scale surveillance system.
Advanced Data Processing and Analytics: While the primary data processing occurs on edge, the cloud offers additional computational resources for deeper data analysis that are not time-sensitive but provide valuable insights, such as trend analysis and behavioral patterns over time.
Backup and Disaster Recovery: The cloud component acts as a vital backup, safeguarding against data loss due to hardware failures or other disruptions at the edge. AWS’s robust infrastructure offers redundancy and ensures data recovery capabilities, enhancing the overall resilience of the surveillance system.
Hardware & Software
We picked the Rockchip RK3399Pro board for its cost-effectiveness and computing power, essential for running computer vision tasks. It has a specialized NPU that efficiently handles AI tasks at a speed of up to 3.0 TOPs, crucial for the immediate analysis required by our client’s security system.
The board’s dual-core Cortex-A72 and quad-core Cortex-A53 setup provides the necessary performance for constant computer vision operations while maintaining energy efficiency. Its ability to work with well-known AI frameworks, like TensorFlow, gave us a range of options for building and implementing our models.
Ultimately, we chose TensorFlow because it has a wide selection of ready-to-use models and works well with the RK3399Pro’s NPU, allowing us to fully use the board’s capabilities. To get the best performance, we tailored TensorFlow models specifically for the NPU, which meant adjusting the models to fit the board’s processing strengths, ensuring our AI solutions ran smoothly within the board’s technical and energy limits.
Value Delivered
