3D Model Extension for Faster Object Pose Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for estimating the position and posture of a target object require the preparation of numerous three-dimensional models, leading to increased calculation costs and times as the number of types increases.
Innovation Solution
An information processing apparatus and method that derive candidate solutions using sensing data and a three-dimensional model, extend the model to improve visibility, and estimate the position and posture without preparing multiple models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If individual precise three-dimensional models are prepared for each type of target object, then the accuracy of position and posture estimation is improved, but the calculation cost and calculation time increase as the number of types increases
Solution Approach 1:
The patent applies universality by using a single generic three-dimensional model that can represent multiple types of target objects. Instead of preparing individual models for each object type, the system uses one model that can be adapted through parameter changes (such as scaling, rotation, and position adjustments) to match different objects. This reduces the number of models from potentially dozens to just one, significantly decreasing calculation time while maintaining estimation accuracy through parameter optimization rather than model differentiation.
2Measurement precision
If individual precise three-dimensional models are prepared for each type of target object, then the accuracy of position and posture estimation is improved, but the device complexity increases
Solution Approach 1:
The patent reduces device complexity by eliminating the need to manage multiple individual three-dimensional models. Instead of storing, managing, and selecting from numerous type-specific models, the system uses a single universal model with adjustable parameters. This simplifies the model management system, reduces storage requirements, and eliminates the complexity of determining which specific model to use for each object type.
Solution Approach 2:
The patent applies parameter changes by modifying the universal model's attributes (such as scale, orientation, and position parameters) to adapt it to different target objects. Rather than changing the model structure itself, the system adjusts numerical parameters to fit different object types, thereby maintaining model simplicity while achieving accurate representations of diverse objects.
3Measurement precision
If individual precise three-dimensional models are prepared for each type of target object, then the accuracy of position and posture estimation is improved, but the calculation cost increases
Solution Approach 1:
The patent reduces calculation cost by using a single universal three-dimensional model instead of multiple type-specific models. This eliminates the need to load, process, and compute with numerous separate models, thereby reducing the computational burden on the processing system. The universal model approach maintains estimation accuracy through parameter adaptation while significantly lowering the energy and computational resources required.
Data Source
AI summary
To realize a technique for estimating the position and posture of a target object without preparing a large number of three-dimensional models, an information processing apparatus (1) includes: an obtaining section (11) that obtains sensing data obtained by a sensor of which a sensing range includes a target object therein; an estimating section (12) that derives one or more candidate solutions to the position and posture of the target object by a first estimating process in which the sensing data obtained by the obtaining section (11) and a three-dimensional model are used; and an extending section (13) that extends the three-dimensional model with respect to each of the one or more candidate solutions, the estimating section (12) estimating the position and posture of the target object by a second estimating process in which one or more extended three-dimensional models obtained by the extending section (13) are used.