Posture Estimation Using Filtered Feature Point Groups
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In information processing systems that estimate the posture of measurement targets from video data, a large number of combinations of measurement targets and feature point groups lead to increased processing load, while reducing the number of types and feature point groups may compromise accuracy.
Innovation Solution
A position measuring program that identifies specific identification information and feature point groups corresponding to a measurement target's position, using stored associations to estimate the posture without matching all combinations, thereby reducing processing load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the number of combinations of storage targets is increased to maintain comprehensive coverage of measurement target types and feature point groups, then the accuracy of posture estimation is improved, but the processing load increases and the time required for estimation becomes longer
Solution Approach 1:
The patent segments the large set of all possible combination matches into two stages: first, a preliminary filtering stage that reduces the candidate combinations based on basic criteria, and second, a detailed matching stage that processes only the reduced set. This segmentation allows comprehensive coverage to be maintained while significantly reducing the actual processing load.
Solution Approach 2:
The patent performs preliminary actions by pre-processing and organizing the storage target data into structured formats with indexed feature point groups and combination metadata. This preliminary organization enables faster retrieval and filtering during runtime, reducing the processing time for posture estimation without sacrificing accuracy.
2Productivity
If the number of types of measurement targets and feature point groups is reduced to decrease processing load, then the processing load is reduced, but the accuracy of estimation of measurement target postures decreases
Solution Approach 1:
The patent implements dynamic processing where the system adaptively adjusts the matching process based on the specific input image data. Rather than statically reducing the number of target types and feature point groups, the system dynamically selects and processes only the relevant combinations needed for the current estimation task, maintaining accuracy while optimizing processing load.
Solution Approach 2:
The patent changes the parameter of combination processing from a static comprehensive approach to a dynamic selective approach. By modifying how combinations are processed (from all combinations to filtered relevant combinations), the system reduces processing load while preserving the necessary accuracy through intelligent parameter selection rather than parameter reduction.
3Measurement precision
If comprehensive matching of all combination types is performed to ensure accurate posture estimation, then the accuracy is maintained, but the time required for estimation becomes longer
Solution Approach 1:
The patent performs preliminary actions by pre-organizing the storage database with indexed and structured combination data, including pre-computed feature point group associations. This preliminary preparation enables rapid filtering and retrieval during runtime, significantly reducing the time required for posture estimation while maintaining comprehensive matching accuracy.
Solution Approach 2:
The patent implements a two-stage matching process where the first stage rapidly skips through and filters out irrelevant combination candidates using basic criteria, and only the relevant candidates proceed to the detailed matching stage. This allows the system to efficiently skip unnecessary processing while ensuring accurate matching for the relevant cases.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A position measuring program including instructions which cause a computer to perform processing including: identifying a first position of a first target in image data; identifying specific identification information and a specific type corresponding to a specific position where a positional relationship with the first position satisfies a first condition, by referring to a first storage device storing a position of each measurement target, identification information on the imaging device having captured the image data, and a type of each target; identifying a specific feature point group corresponding to the specific identification information and the specific type, by referring to a second storage device storing the identification information on an imaging device having captured the image data, the type of each target, and a feature point group of each target in the image data; and estimating a posture of the first target from the specific feature point group.