Robot Object Recognition via Property-Estimated Image Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object recognition methods in robots often struggle with categorizing objects with insufficient data in their libraries, leading to low confidence scores and inaccurate identification.
Innovation Solution
A method where a robot determines a preliminary category of an object with a confidence score, estimates properties if the score is below a learning threshold, and searches a supplemental image collection for data to supplement its library, using the preliminary category and estimated properties to improve recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the robot uses a vision-based object recognition method with a library of reference images, then it can categorize objects by comparing features, but the library may have insufficient data regarding particular types or categories of objects such that the robot cannot categorize a target object with a high degree of confidence
Solution Approach 1:
The robot performs preliminary actions by estimating properties (color, shape, size, texture) of the target object before searching for supplemental image data. This preliminary property estimation guides the search query to find relevant reference images that can improve categorization confidence.
Solution Approach 2:
The system implements feedback by using the confidence score from initial categorization attempts to determine whether supplemental data collection is needed. When confidence is low, the system searches for and adds supplemental image data to the library, then re-evaluates categorization with the enhanced library.
2Measurement precision
If the robot searches for and collects more image data to improve recognition accuracy, then categorization confidence improves, but this increases data collection and storage requirements
Solution Approach 1:
The robot applies local quality by estimating specific properties (color, shape, size, texture) of the target object and using these localized property estimates to construct targeted search queries. This approach focuses data collection on specific property dimensions rather than collecting all possible image data.
Solution Approach 2:
The system performs partial action by collecting only the specific supplemental image data needed to improve confidence for particular object categories with insufficient data, rather than collecting comprehensive data for all categories. This selective approach reduces overall data storage requirements.
Data Source
AI summary
Methods for recognizing a category of an object are disclosed. In one embodiment, a method includes determining, by a processor, a preliminary category of a target object, the preliminary category having a confidence score associated therewith, and comparing the confidence score to a learning threshold. If the highest confidence score is less than the learning threshold, the method further includes estimating properties of the target object and generating a property score for one or more estimated properties, and searching a supplemental image collection for supplemental image data using the preliminary category and the one or more estimated properties. Robots programmed to recognize a category of an object by use of supplemental image data are also disclosed.