Symmetry-Aware Posture Prediction for Projection Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object posture prediction technologies face inefficiencies when dealing with symmetric objects, as they result in longer computational times and unstable predictions due to overlapping feature data and multiple candidate points, which are not effectively handled by current methods.
Innovation Solution
The approach involves establishing a partial region on the object that satisfies specific conditions to prevent overlapping feature data storage and using machine learning to predict object posture, considering the symmetry of the object to handle multiple similar hypotheses robustly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If feature quantities are extracted from all points on the object for posture prediction, then the prediction accuracy is improved, but the computational time increases significantly for symmetric objects
Solution Approach 1:
The patent extracts only the necessary feature quantities from representative points rather than all points on the object. By identifying and extracting features from a subset of points that are sufficient for accurate posture prediction, the system reduces computational load while maintaining prediction accuracy, directly addressing the contradiction between accuracy and time.
Solution Approach 2:
The patent introduces asymmetry handling by detecting whether an object has symmetric properties and adjusting the feature extraction process accordingly. When symmetry is detected, the system uses this information to reduce the number of points needing feature extraction, as symmetric points would produce identical feature quantities anyway, thus reducing computational time without losing prediction accuracy.
2Adaptability or versatility
If multiple candidate points are considered for symmetric objects, then the prediction covers more possibilities, but the matching process becomes longer and less stable
Solution Approach 1:
The system detects symmetric properties of objects and uses this information to adjust the matching process. By identifying symmetry, the system can reduce the number of candidate points to consider during matching, as symmetric points would generate identical feature quantities, thereby reducing matching time while maintaining comprehensive prediction coverage.
Solution Approach 2:
The patent changes the approach to handling symmetric objects by introducing a symmetry detection parameter that modifies the feature extraction and matching process. When symmetry is detected, the system adjusts which points are processed and how matching is performed, reducing computational complexity while maintaining prediction versatility.
3Adaptability or versatility
If overlapping feature data is stored in the dictionary, then all symmetric points are represented, but the data redundancy increases and processing efficiency decreases
Solution Approach 1:
The patent addresses data redundancy by detecting object symmetry and adjusting the feature extraction process accordingly. When symmetry is detected, the system extracts features only from representative points rather than all symmetric points, eliminating redundant entries in the dictionary while maintaining complete feature representation, thus improving processing efficiency.
Solution Approach 2:
The system introduces a symmetry parameter that changes the feature extraction process. When symmetry is detected, the system modifies which points are processed and how features are stored, reducing data redundancy in the dictionary while preserving complete object representation, thereby improving productivity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present technology relates to an information processing apparatus, an information processing method, and a program that are capable of easily predicting the posture of an object. An information processing apparatus according to an aspect of the present technology specifies, on the basis of learned data used in specifying corresponding points, obtained by performing learning using data of a predetermined portion that has symmetry with respect to other portions of an entire model that represents an object as a recognition target, second points on the model included in an input scene that correspond to first points on the model, as the corresponding points, and predicts the posture of the model included in the scene on the basis of the corresponding points. The present technology is applicable to an apparatus for controlling a projection system to project images according to projection mapping.