Position-Invariant Feature Extraction for Place Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional PIRF methods fail to distinguish between static and kinetic local feature values, leading to inaccurate place estimation and reduced calculation speed due to the inclusion of moving objects as invariant features.
Innovation Solution
The development of a feature value extraction method, known as Incremental Center of Gravity Matching (ICGM), which separates position-invariant feature values by comparing the geometrical positional relations between feature points in successive images, using a position-invariant feature value extraction unit to identify features with unchanged positions over time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If PIRF extracts all local feature values that appear in successive frames as invariant feature values, then the quantity of feature values increases, but place estimation accuracy deteriorates due to inclusion of kinetic features
Solution Approach 1:
The patent segments feature values into two distinct categories: position-invariant features (static objects) and position-variant features (kinetic objects). This segmentation is achieved by comparing feature positions across multiple successive frames and identifying features that maintain consistent positions. By separating these categories, the system extracts only position-invariant features for place estimation, eliminating the negative impact of kinetic features while preserving sufficient quantity of valid features.
Solution Approach 2:
The patent extracts and isolates position-invariant feature values from the mixture of all local feature values. Through position comparison across successive frames, the system identifies and extracts only those features whose positions remain unchanged, separating them from kinetic features. This extraction process ensures that only relevant static features are used for place estimation, improving accuracy while maintaining adequate feature quantity.
2Device complexity
If PIRF treats all successive local feature values as invariant, then feature extraction complexity is reduced, but calculation speed deteriorates due to processing of irrelevant kinetic features
Solution Approach 1:
The patent performs preliminary position comparison and classification of features before the main place estimation calculation. By pre-identifying and filtering position-invariant features from kinetic features in advance, the system reduces the computational burden during subsequent place estimation operations. This preliminary filtering action eliminates irrelevant kinetic features early in the processing pipeline, thereby improving overall calculation speed without significantly increasing extraction complexity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A feature value extraction apparatus, a method, and a program capable of extracting local feature values whose positions are unchanged, and a place estimation apparatus, a method, and a program equipped with them are provided. A place estimation apparatus (10) performs a place estimation process by using position-invariant feature values extracted by a feature value extraction unit (11). The feature value extraction unit (11) includes local feature value extraction means (21) for extracting local feature values from each of input images formed from successively-shot successive images, feature value matching means (22) for obtaining matching between successive input images based on the extracted local feature values, corresponding feature value selection means (23) for selecting matched feature values as corresponding feature values, and position-invariant feature value extraction means (24) for obtaining position-invariant feature values based on the corresponding feature values. The position-invariant feature value extraction means (24) extracts, from among the corresponding feature values, corresponding feature values whose position change is equal to or less than a predetermined threshold as the position-invariant feature values.