Object Mapping Device Using Overlapping Time Segments for Accurate Entity Association
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for associating objects in multiple contents, such as images, face challenges when objects change over time, leading to inaccurate grouping, especially due to factors like aging, weight gain/loss, or feature changes that deviate from statistical averages.
Innovation Solution
An object associating device that categorizes objects into groups based on generation time information, dividing the time period into overlapping segments where objects are similar, and associates objects across these segments if a predetermined number is met, thereby minimizing the impact of changes within each segment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If clustering is performed simply based on feature values of faces, then the clustering process is simple and fast, but images of the same person captured at different life stages are separated into different clusters
Solution Approach 1:
The patent segments the time period into multiple sub-periods and performs clustering separately for each sub-period. This allows the system to capture facial features at different life stages while maintaining the ability to associate them through temporal relationships, thereby improving association accuracy without sacrificing too much processing efficiency.
Solution Approach 2:
The patent introduces a temporal dimension by dividing the time period into sub-periods and performing clustering in both time and feature spaces. This multi-dimensional approach enables the system to handle facial changes over time while maintaining accurate person association across different life stages.
2Adaptability or versatility
If statistical data is used to judge sameness considering growth and aging, then the judgment becomes more comprehensive, but accuracy decreases for individuals whose changes deviate from statistical averages
Solution Approach 1:
By segmenting the time period into smaller sub-periods, the patent reduces the magnitude of changes within each segment. This allows the system to make more accurate local judgments about facial similarity while still considering overall temporal patterns, thereby improving precision for individuals with non-typical growth patterns.
Solution Approach 2:
The patent performs clustering for each sub-period separately rather than making a single comprehensive judgment. This partial action approach allows the system to accumulate evidence across multiple time points, improving accuracy for individuals whose changes deviate from statistical averages by considering multiple intermediate states.
3Measurement precision
If the time period is divided into overlapping sub-periods for clustering, then association accuracy improves by considering changes within shorter periods, but the complexity of the clustering process increases
Solution Approach 1:
The patent divides the time period into overlapping sub-periods and performs clustering for each segment. This segmentation allows the system to handle facial changes more accurately by focusing on smaller time intervals where changes are more gradual, while the overlapping structure provides continuity for association across periods.
Solution Approach 2:
The patent merges the results from multiple sub-period clusterings by identifying common objects across overlapping periods. This combining approach allows the system to leverage information from all sub-periods while maintaining the benefits of localized clustering, thereby improving accuracy without proportionally increasing complexity.
Data Source
AI summary
A device stores therein contents and time information pieces each indicating time at which the corresponding content was generated; categorizes, into first and second groups, objects included in contents corresponding to time information pieces indicating times within a first period and a second period that partially overlap, the first group composed of objects that are included in contents corresponding to time information pieces indicating times within the first period and are similar to each other, and the second group composed of objects that are included in contents corresponding to time information pieces indicating times within the second period and are similar to each other; and when a number of objects belonging to both the first and second groups is greater than or equal to a predetermined number, associates the objects belonging to the first group with the objects belonging to the second group as objects indicating a same entity.