Feature Point Selection for Position Orientation Calculation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for measuring the position/orientation of an image capturing apparatus in the presence of moving objects suffer from low precision and stability due to incorrect association of feature points, leading to inaccurate calculations.
Innovation Solution
An information processing apparatus and method that selectively uses feature points with high availability for position/orientation calculations by associating detected features with held features and excluding those near moving objects, based on their stability and precision indices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If feature points are used for position/orientation calculation without filtering, then the calculation can be performed with all available data, but the precision and stability deteriorate when moving objects are present in the scene
Solution Approach 1:
The system performs preliminary classification of feature points into moving and stationary categories before position/orientation calculation. By pre-identifying stationary feature points through temporal comparison and motion analysis, the system ensures that only reliable features are used in the calculation, thereby maintaining high precision without requiring complex real-time filtering during the calculation process itself.
Solution Approach 2:
The system applies different quality criteria to different feature points based on their motion characteristics. Stationary feature points are selected for calculation while moving feature points are excluded. This local differentiation in feature point quality assessment allows the system to maintain high measurement precision by using only suitable features, rather than applying a uniform selection criterion to all features.
2Reliability
If all detected feature points are used in position/orientation calculation, then the quantity of data is maximized, but the reliability deteriorates due to incorrect association with moving objects
Solution Approach 1:
The system extracts and separates stationary feature points from moving feature points by analyzing temporal changes in feature positions. By taking out only the stationary feature points that remain consistent across multiple frames and excluding moving feature points that show temporal variation, the system maintains high calculation reliability while using an optimized subset of feature points rather than all detected features.
Solution Approach 2:
The system uses feedback from temporal comparison of feature points across multiple frames to identify and exclude moving objects. By continuously monitoring feature point consistency over time and using this feedback to adjust the selection of feature points for calculation, the system dynamically maintains high reliability by adapting to changing scene conditions and excluding newly detected moving objects.
Data Source
AI summary
An information processing apparatus inputs a plurality of detected features that are detected from an image or three-dimensional data, holds a plurality of features of a physical space as a plurality of held features in association with availabilities in a calculation of a position/orientation of an apparatus that obtained the image or the three-dimensional data, decides spatial locations of the held features in the image or the three-dimensional data, and decides, based on the availabilities of the held features existing in a vicinity of one detected feature out of the plurality of detected features at the locations, a state of usage of the one detected feature in the position/orientation calculation.