Object Recognition Apparatus Using Dynamic Data Store Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Computer vision systems face challenges in identifying objects when pre-specified object templates are not available in their data stores, as they rely on matching features and elements with stored templates, and lack access to additional object data stores for identification.
Innovation Solution
The system captures visual data, compares candidate objects with object templates, and if no match is found, it uses detected features and elements to search associated and non-associated data stores, including remote sources via networks, to identify objects by accessing and integrating additional object templates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If the system relies on matching features with pre-specified object templates in proprietary data stores, then identification speed is improved, but identification completeness deteriorates when objects are not in the data store
Solution Approach 1:
The system dynamically adjusts its identification approach based on whether template matching succeeds or fails. When template matching fails, the system transitions to feature-based search in non-proprietary data stores, making the identification process adaptive rather than static.
Solution Approach 2:
The patent introduces non-proprietary data stores as intermediaries between the proprietary data store and the identification process. When proprietary templates are insufficient, the system queries external data stores to supplement object identification, acting as a mediator to bridge the gap between limited proprietary data and comprehensive object recognition.
2Reliability
If the system accesses only proprietary data stores, then data security and control are improved, but object recognition capability deteriorates when objects are not in the proprietary data store
Solution Approach 1:
The patent segments the data store access into two distinct parts: proprietary data stores for secure, controlled access and non-proprietary data stores for supplementary object identification. This segmentation allows the system to maintain data control over proprietary information while still accessing external resources when needed.
Solution Approach 2:
The system adds another dimension to data store access by incorporating non-proprietary data stores alongside proprietary ones. This dimensional expansion allows the system to maintain data control in the proprietary dimension while gaining enhanced object recognition capability through the non-proprietary dimension.
3Reliability
If the system searches multiple non-proprietary data stores, then object identification completeness is improved, but system complexity and processing time deteriorate
Solution Approach 1:
The system performs preliminary template matching against proprietary data stores before querying non-proprietary data stores. This preliminary action filters out objects that can be quickly identified, reducing the number of objects that require complex multi-store searching and thereby reducing overall system complexity.
Solution Approach 2:
The patent applies different identification strategies to different objects based on local characteristics. Objects with distinctive features are matched against proprietary templates first, while objects requiring more comprehensive search are then queried in non-proprietary data stores, optimizing the balance between completeness and complexity.
Data Source
AI summary
A method of identifying an object within an environment is described for capturing visual data associated with the environment and comparing data defining a candidate object in the environment with an object data store storing object templates providing data pertaining to one or more objects; responsive to matching the data defining the candidate object in the environment and an object template, identifying the candidate object; and responsive to failing to match data representing the candidate object in the environment with an object template, identifying one or more object identifiers disposed in the environment to define identifier search data, and using the identifier search data to interrogate a further object data store to identify the candidate object.