Tracking Inference Learning from Posture-Linked Training Data

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing tracking technologies face accuracy issues due to errors in posture estimation, which affect the tracking process of objects in time series images.

Innovation Solution

A learning device and method that utilize tracking training data to associate position and posture information with identification information, learning an inference engine to infer correspondence relations between images, using neural networks and other learning models to enhance tracking accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If tracking is performed based on posture information from time series images, then tracking capability is achieved, but tracking accuracy deteriorates when posture estimation errors occur

Engineering Contradiction:

Improvetracking accuracyVSAvoidposture estimation accuracy

Core Design Contradiction:

ReliabilityVSMeasurement precision

Solution Approach 1:

The system performs preliminary association between posture information and identification information before tracking. By pre-linking posture data with unique identifiers across multiple images, the system creates a robust tracking foundation that is less sensitive to posture estimation errors during actual tracking operations

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system introduces identification information as an intermediary element between posture information and tracking results. This mediator allows the system to track objects reliably even when posture estimation varies, as the identification information provides a stable reference that decouples tracking accuracy from posture estimation precision

Inventive Principle:

Principle #24Intermediary (Mediator)

2Adaptability or versatility

If posture information is estimated from images to enable tracking, then tracking functionality is achieved, but tracking reliability deteriorates due to estimation errors

Engineering Contradiction:

Improvetracking functionalityVSAvoidtracking reliability

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system uses identification information as a feedback mechanism to maintain tracking reliability. By continuously referencing the pre-established associations between posture information and identification information, the system can correct or compensate for posture estimation errors, ensuring consistent and reliable tracking performance

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS12417546B2Learning device, learning method, tracking device, and storage medium

Publication Date: 2025.09.16 NEC CORP

AI summary

A learning device 1X includes an acquisition means 15X, an estimation result matching means 16X, and a learning means 18X. The acquisition means 15X acquires tracking training data in which first and second training images in time series, tracking target position information regarding position or posture of a tracking target shown in each first and second training images, and identification information of the tracking target are associated. The estimation result matching means 16X compares the tracking target position information with posture information indicating posture of the tracking target estimated from the first and second training images and associates the posture information with the identification information. The learning means 18X learns an inference engine, which infers correspondence information indicating correspondence relation of the tracking target between the training images when information based on the posture information is inputted, the correspondence information based on the posture information and the identification information.