Pose Determination via Line-Based Texture Map Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing human action classification techniques using neural networks struggle to accurately determine specific poses due to enhanced focus on body context rather than pose features.
Innovation Solution
An image forming apparatus and method that creates a feature extracted image by drawing lines between key points in a target's pose, forming a texture map that emphasizes spatial pose features, improving pose determination accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If a human shaped image painted with certain color on each part of the human body is used, then the body context is enhanced, but the specific pose feature is weakened leading to inaccurate pose classification
Solution Approach 1:
The patent segments the human body representation into discrete key points and connects them with lines to form a skeletal structure. This segmentation separates the pose-specific geometric information from the body context, allowing the neural network to focus on pose determination while the line-based representation maintains the spatial relationships between body parts.
Solution Approach 2:
The patent applies local quality by using different visual representations for different purposes: key points mark specific body locations, while lines connecting them emphasize the spatial relationships and pose geometry. This local differentiation ensures that pose-critical features are highlighted without being overwhelmed by overall body context.
2Measurement precision
If lines are drawn between key points with ends extended to image boundary, then the spatial pose features are emphasized, but the device complexity increases
Solution Approach 1:
The patent performs preliminary action by pre-defining the line drawing pattern before neural network processing. The lines are systematically drawn between predetermined key points with ends extended to image boundaries, creating a standardized feature representation that simplifies subsequent processing while emphasizing pose geometry.
Solution Approach 2:
The patent transitions from a 2D image space to a structured feature space by drawing lines that extend to image boundaries. This dimensional transformation creates a more robust pose representation that captures spatial relationships more effectively, adding structural information that aids pose determination.
Data Source
AI summary
In an image forming apparatus (10), a feature extracted image forming unit (11) includes a line-drawing unit (11A). The line-drawing unit (11A) draws in an image to be processed, for each of a plurality of line-draw groups, a line that passes all the key points included in each of the line-draw groups and has at least one end that is extended to an end of the image to be processed, thereby forming a texture map. Each of the line-draw groups includes at least two key points included in the plurality of predetermined key points.