Edge-Cloud Visual Computing With Parallel Video, Feature, and Model Streams
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional video surveillance systems face challenges in managing the rapid expansion of data, leading to inefficient processing, resource waste, and inability to meet real-time processing requirements due to inflexible system architecture and limited backend capacity.
Innovation Solution
A scalable visual computing system with a parallel data transmission architecture involving a front-end device, edge service, and cloud service, where compressed video and compact feature streams are transmitted in real-time, and model streams are updated episodically, enabling dynamic model deployment and real-time data analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a traditional video surveillance system employs a labor-division model with cameras for acquisition and backend servers for processing, then the system structure is simple, but the system bandwidth pressure increases and backend processing capacity becomes insufficient
Solution Approach 1:
The patent divides the video surveillance system into three distinct layers: front-end devices (cameras with embedded processing), edge services (regional processing nodes), and cloud services (centralized management). This segmentation distributes processing tasks across multiple levels, preventing backend overload while maintaining manageable system complexity through clear functional boundaries at each layer.
Solution Approach 2:
The patent introduces a new dimensional architecture by adding the edge computing layer between front-end and cloud backends. This creates a three-dimensional processing hierarchy rather than traditional two-dimensional (acquisition-processing) structure, enabling spatial distribution of computing tasks and improving overall system throughput without linearly increasing backend complexity.
2Ease of manufacture
If cameras are configured with fixed functions in a traditional system, then the system is easy to deploy, but the adaptability to different application services is poor
Solution Approach 1:
The patent implements dynamic adaptability by enabling front-end devices to dynamically load and switch algorithm models based on application requirements. The system can dynamically adjust processing functions, data transmission formats, and analysis algorithms without physical reconfiguration, allowing the same hardware to serve multiple applications while maintaining simple initial deployment.
Solution Approach 2:
The patent creates universal front-end devices that can perform multiple functions through software-defined capabilities. A single camera device can serve different application services (surveillance, recognition, analysis) by loading appropriate algorithm models, eliminating the need for dedicated hardware for each function while keeping deployment straightforward.
3Loss of information
If a large amount of video data is transmitted to the backend for processing, then comprehensive analysis is achieved, but the data transmission pressure on system bandwidth increases significantly
Solution Approach 1:
The patent extracts and processes critical information at the edge computing layer before transmitting to the cloud. Instead of sending raw video data, the system extracts key features, detection results, and structured data at the edge, transmitting only essential information to the cloud backend. This maintains analysis completeness while dramatically reducing bandwidth consumption.
Solution Approach 2:
The patent performs preliminary processing and filtering at the front-end and edge layers before data reaches the cloud backend. Video data undergoes preliminary analysis, compression, and selection of relevant frames/events at lower layers, so that only pre-processed, high-value data requires cloud transmission, reducing overall data volume while preserving necessary analytical information.
4Device complexity
If video data is processed in a centralized backend system, then data management is simplified, but the real-time processing capability is insufficient for massive data
Solution Approach 1:
The patent segments data management across multiple hierarchical levels: front-end devices manage local data collection and initial processing, edge services handle regional data aggregation and real-time processing, and cloud services perform centralized long-term storage and macroscopic analysis. This segmentation enables real-time processing at edge levels while maintaining simplified centralized management at cloud level for non-time-critical operations.
Data Source
AI summary
A visual computing system is disclosed. The visual computing system may include a front-end device, an edge service and a cloud service which are in communication connection, the front-end device is configured to output compressed video data and feature data, the edge service is configured to store the video data, and converge the feature data, transmit various types of data and control commands, and the cloud service is configured to store algorithm models used to support various applications, and return a model stream according to a model query command, realizing a data transmission architecture with multiple streams of video stream, feature stream, and model stream in parallel, and a system architecture of end, edge, and cloud collaboration.