Rinegan: A Scalable Image Processing Architecture for Large Scale Surveillance Applications

Area of Science:

• Robotics and Computer Vision
• Intelligent Surveillance Systems

Background:

• Single robot cameras offer limited field of view and processing power for large-scale surveillance.
• Tracking suspects in vast areas necessitates cooperation between fixed cameras and mobile robots, increasing resource demands.

Purpose of the Study:

• To develop a scalable architecture for optimizing image processing efficacy and response rates in intelligent robot visual surveillance.
• To address the challenges of bandwidth consumption and computational load in robot-assisted large-scale monitoring.

Main Methods:

• Implemented a micro-service architecture for flexibility and orchestration.
• Deployed lightweight pre-processing and object detection on gateway-side devices to reduce bandwidth.
• Utilized cloud-side servers for specific suspect identification, receiving only recognized data.
• Developed a prototype system named Rinegan for evaluation.

Main Results:

• The Rinegan system demonstrated improved effectiveness and efficacy in image processing tasks.
• The architecture successfully minimized bandwidth consumption by processing data at the gateway.
• Cloud-side analysis of recognized data enabled efficient suspect identification and tracking coordination.

Conclusions:

• The proposed scalable architecture enhances the capabilities of intelligent robots in surveillance applications.
• Rinegan provides an effective solution for optimizing image processing and response times in complex visual monitoring scenarios.
• Micro-service orchestration ensures flexibility and efficiency in distributed intelligent surveillance systems.