Edge Video Search Using Distributed Storage and Attribute Indexing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Internet-of-Things (IoT) applications face bandwidth and backhaul limitations, as well as storage constraints, making it challenging to access and retrieve visual data captured by connected devices efficiently, particularly in systems like intelligent driver monitoring systems and smart surveillance.
Innovation Solution
The implementation of edge computing methods that involve extracting visual data attributes, vectorizing them, and indexing them in a schema-less database, allowing for distributed video storage and search. This includes context-specific model selection and joint client-cloud processing, variable rate deep neural networks, and data caching to manage bandwidth and storage effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If all visual data is transmitted to centralized servers for storage and search, then data accessibility is improved, but bandwidth consumption increases and storage limitations are exceeded
Solution Approach 1:
The system segments visual data into extracted attributes (features, metadata) and full data components. Only the essential attributes are transmitted to centralized servers, while full data remains distributed at edge devices. This segmentation enables data accessibility for search operations without requiring transmission of complete visual datasets, thereby reducing bandwidth consumption significantly.
Solution Approach 2:
The system extracts key visual attributes (features, metadata, descriptors) from complete visual data at edge devices before transmission. This extraction process isolates and transmits only the essential information needed for search and identification purposes, leaving the bulk data stored locally. This resolves the contradiction by providing data accessibility through extracted attributes while minimizing bandwidth consumption.
2Quantity of substance
If visual data is stored at edge devices, then storage capacity is improved, but data search and retrieval efficiency deteriorates
Solution Approach 1:
The system introduces extracted visual attributes (features, metadata, descriptors) as intermediaries between stored visual data and search queries. These attributes serve as searchable proxies that enable efficient data retrieval without requiring access to the full visual datasets. The intermediaries are stored centrally and can be queried rapidly, maintaining search efficiency while allowing bulk data to remain distributed at edge devices.
Solution Approach 2:
The system creates simplified copies of visual data in the form of extracted attributes (features, metadata, descriptors). These attribute copies contain essential information for search and identification but occupy minimal storage space. The copies are transmitted to centralized servers where they enable efficient search operations across distributed data without requiring the actual visual data to be centralized.
3Measurement precision
If visual data is processed centrally, then processing accuracy is improved, but latency increases due to data transmission
Solution Approach 1:
The system performs preliminary processing at edge devices by extracting visual attributes (features, metadata, descriptors) before data transmission. This preliminary extraction prepares the data in advance, creating ready-to-use searchable attributes that can be quickly transmitted and processed centrally. The preliminary action reduces the complexity of subsequent central processing and minimizes transmission latency by sending only essential extracted data rather than complete visual datasets.
4Loss of information
If bandwidth is increased to transmit all visual data, then data accessibility is improved, but system cost and complexity increase
Solution Approach 1:
The system extracts and transmits only essential visual attributes (features, metadata, descriptors) rather than complete visual data. This extraction approach provides sufficient data accessibility for search and identification operations while dramatically reducing the bandwidth requirements. The simplified data transmission reduces network infrastructure complexity and costs while maintaining the core functionality of distributed visual data access.
Data Source
Figure 1A
Figure 1B
Figure 1C
AI summary
Systems and methods are provided for distributed video storage and search with edge computing. The method may comprise caching a first portion of data on a first device. The method may further comprise determining, at a second device, whether the first device has the first portion of data. The determining may be based on whether the first piece of data satisfies a specified criterion. The method may further comprise sending the data, or a portion of the data, and/or a representation of the data from the first device to a third device.