Distributed Video Analysis Object Detection Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video surveillance systems face issues with increased data traffic and high costs due to the need for high-performance processing of image data, which limits bandwidth and requires frequent updates.
Innovation Solution
The system separates object detection and analysis functions, with object detection occurring near cameras and sending only relevant data for analysis, using techniques like neural networks and deep learning to identify objects of interest, reducing the amount of data transmitted over networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If surveillance cameras send all image data over the network for processing, then object analysis can be performed accurately, but network bandwidth is consumed and performance problems occur
Solution Approach 1:
The video surveillance system is divided into multiple processing nodes: edge devices (cameras, NVRs) perform preliminary object detection and filtering, while a central video analytics system performs detailed analysis. This segmentation allows only relevant image data containing detected objects to be transmitted over the network, significantly reducing data traffic while maintaining recognition accuracy.
2Measurement precision
If high-performance video analytics systems are used to improve analysis accuracy, then object detection and recognition improve, but system cost increases
Solution Approach 1:
The system distributes processing capabilities across multiple nodes with different performance levels. Edge devices perform lightweight object detection, while the central analytics system performs sophisticated analysis. This allows the use of advanced algorithms without requiring every component to be high-performance, reducing overall system cost.
Solution Approach 2:
The patent introduces an intermediary filtering layer at edge devices that preprocesses video data before transmission. This intermediary performs initial object detection and filters out irrelevant frames, reducing the burden on the central high-performance analytics system and allowing cost optimization.
3Measurement precision
If image data is transmitted at high resolution for better analysis, then object recognition improves, but network bandwidth consumption increases
Solution Approach 1:
Object detection and filtering are performed preliminarily at edge devices before image data is transmitted to the central analytics system. This preliminary action identifies which frames contain objects of interest, allowing only those frames to be transmitted at full resolution, while other frames are transmitted at lower resolution or not at all.
Data Source
AI summary
A system and method for distributed analysis of image data in a video surveillance system can be deployed in households, businesses, and within corporate entities, in examples. The image data are processed on local video analytics systems located within the networks of the businesses or on remote video analytics systems hosted by a cloud service. To limit the data traffic imposed by the image data on the network, the system divides the image data processing into separate object detection and analysis functions. The system can also integrate the object detection function within the surveillance cameras or on a local gateway. This can significantly reduce the data traffic sent over networks as compared to current video surveillance systems and methods since only image data containing object of interest needs to be sent.