Surveillance System Server Camera Load Distribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current surveillance systems face challenges in efficiently processing high-definition image data, leading to increased network traffic and processing loads on servers when performing image recognition tasks using deep learning, as the arithmetic processing capability of cameras is insufficient and the transfer of high-quality image data to servers results in communication inefficiencies and delays.
Innovation Solution
A surveillance system where the server distributes the process of learning parameters among multiple cameras based on their processing capabilities, allowing each camera to execute specific tasks and share learning results, thereby reducing network traffic and server processing loads.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the server performs image recognition process on high-definition captured image data, then detection accuracy is improved, but network traffic increases and communication delay occurs
Solution Approach 1:
The patent segments the image recognition process by dividing responsibilities between cameras and servers. Cameras perform preliminary processing including object detection and parameter learning, while the server handles only the aggregation and final recognition. This segmentation reduces the amount of data transmitted over the network while maintaining detection accuracy.
2Measurement precision
If the server performs image recognition process on high-definition captured image data, then detection accuracy is improved, but communication delay increases
Solution Approach 1:
The patent implements preliminary action by having cameras perform object detection and parameter learning before transmitting data to the server. This preliminary processing at the camera end reduces the time required for server-side processing and minimizes communication delay, as only essential processed data needs to be transmitted.
3Measurement precision
If learning process is centralized on the server, then parameter accuracy is improved, but server processing load increases
Solution Approach 1:
The learning process is segmented between cameras and the server. Cameras perform local parameter learning using captured image data, and the server aggregates these learned parameters to generate final recognition parameters. This distribution reduces the server's processing load while maintaining parameter accuracy through collaborative learning.
4Quantity of substance
If cameras perform image recognition process locally, then network traffic is reduced, but processing capability requirements increase
Solution Approach 1:
The patent applies partial action by having cameras perform only the necessary preliminary processing (object detection and parameter learning) rather than complete image recognition. This approach reduces network traffic while avoiding the need for cameras to have excessive processing capabilities, as the server complements the camera's partial processing.
Data Source
AI summary
In a surveillance system, a server and a plurality of cameras provided in a surveillance area are communicably connected to each other. The server includes a table memory that retains information on free resources of each of cameras. The server determines for each camera, a process to be executed by the camera, based on the information on the free resources of the camera, and transmits an instruction to execute the determined process to each camera. Each of the cameras executes a process corresponding to the instruction to execute, based on an instruction to execute the process transmitted from the server.