Feet Position Estimation Using Head Reference Images

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing person recognition systems, such as those described in Patent Literature 1, lack accuracy in detecting the feet position of a person when it is hidden, which hinders the generation of accurate trajectory information for security enhancements in public facilities.

Innovation Solution

An information processing device comprising a reference image generation unit, a comparative image generation unit, an adjustment unit, and a feet estimation unit, which generates and compares reference and comparative images to adjust the head position in the tracking image, allowing for accurate estimation of the feet position using machine learning and deep learning techniques.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If a filter is used to detect the head of a target person from an image, then the head detection process is simplified, but the accuracy of head detection is insufficient

Engineering Contradiction:

Improvehead detection processVSAvoidhead detection accuracy

Core Design Contradiction:

Ease of operationVSMeasurement precision

Solution Approach 1:

The patent creates a reference image by extracting and copying the head portion from a whole-body image. This reference image is then used to guide the detection process in tracking images, allowing the system to accurately identify the head position by comparing against the known reference rather than relying solely on filters.

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The reference image serves as an intermediary element between the whole-body image and the tracking image. By introducing this intermediate reference, the system can bridge the gap between having complete body information and needing accurate head detection in situations where the full body is not visible.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Adaptability or versatility

If the feet position is hidden in the tracking image, then the tracking image can capture important movement information, but the accuracy of feet position estimation deteriorates

Engineering Contradiction:

Improvetracking image capture capabilityVSAvoidfeet position estimation accuracy

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent segments the image into a head portion (extracted from the whole-body image) and the rest of the body portion (from the tracking image). By separating these segments, the system can use the known head position from the reference to infer the feet position, even when the feet are hidden in the tracking image.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system transitions from two-dimensional image data to three-dimensional spatial reasoning by using the known head position and assumed body height to calculate the feet position. This dimensional approach allows the system to infer hidden information based on geometric relationships.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

3Measurement precision

If multiple assumed heights are used to generate movement trajectories, then the height estimation accuracy is improved, but the computational complexity increases

Engineering Contradiction:

Improveheight estimation accuracyVSAvoidcomputational complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent performs preliminary action by extracting the head portion from the whole-body image and creating a reference image before the actual tracking process. This pre-extraction eliminates the need to test multiple assumed heights during tracking, as the head position is already known from the reference image.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system extracts only the necessary information (head position) from the whole-body image to create the reference image. This extraction approach eliminates the need to process multiple assumed heights, reducing computational complexity while maintaining accuracy.

Inventive Principle:

Principle #2Taking out (Extraction)

Data Source

PatentUS12067735B2Information processing device, estimation method, and nontransitory computer readable medium for estimating a position of a target

Publication Date: 2024.08.20 NEC CORP

AI summary

An information processing device includes a reference image generation unit that generates a reference image by using a partial area including a head of a tracking target in a whole-body image of the tracking target, a comparative image generation unit that generates a comparative image to be used for comparison with the reference image by using the partial area containing the head of the tracking target in a tracking image of the tracking target where feet are not contained, an adjustment unit that compares the reference image with the comparative image, and adjusts a position of the head of the tracking target in the comparative image, and a feet estimation unit that estimates a position of the feet of the tracking target in the tracking image by using the comparative image where the position of the head of the tracking target is adjusted.