2D Image Sequence to 3D Body Model Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for body scanning, such as using full body scanners or 2D image analysis, are costly and complex, making them unsuitable for mass deployment in stores or consumer use, and fail to provide accurate 3D measurements due to insufficient information from single or multiple 2D views.
Innovation Solution
A system and method that captures a sequence of 2D images from a single static camera to derive accurate body size measures by integrating 2D shape descriptors into 3D size measures and shape elements, using image analysis to track user movement and posture, and convert silhouette data into 3D models for augmented reality visualization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full body scanners (laser scanners or array of depth cameras) are used to capture 3D body data, then measurement precision is improved, but device complexity and cost increase significantly
Solution Approach 1:
The patent uses a 2D camera to capture multiple images of the body from different angles and reconstructs a 3D body model by processing these 2D images. This approach creates a virtual copy of the 3D body shape from 2D projections, avoiding the need for complex 3D scanning hardware while achieving accurate 3D measurements through computational methods
Solution Approach 2:
The patent replaces the mechanical 3D scanning system (laser scanners or depth camera arrays) with a simpler 2D camera system that captures images from multiple angles. The 3D reconstruction is achieved through computational processing of the 2D images rather than direct 3D optical measurement, substituting mechanical complexity with computational algorithms
2Measurement precision
If multiple 2D images from different views are used to extract body measurements, then measurement precision is improved, but device complexity increases due to controlled turntable or on-body markers
Solution Approach 1:
The patent enables the user to capture their own body images from multiple angles using a simple 2D camera without requiring external pose control systems. The user can freely move and position themselves while the system automatically processes the images to extract 3D body measurements, eliminating the need for controlled turntables or on-body markers
Solution Approach 2:
The patent allows the user to dynamically change their pose and position freely while capturing images, rather than being constrained to fixed poses by controlled turntables or markers. The system processes these dynamic images to extract accurate 3D body measurements, accommodating natural user movement without requiring rigid pose control
3Ease of operation
If approximations (e.g. elliptical cross section) are used to compute body circumference from 2D images, then ease of operation is improved, but measurement precision deteriorates
Solution Approach 1:
The patent transitions from 2D image analysis to 3D body model reconstruction by processing multiple 2D images from different angles. This dimensional transformation allows the system to compute accurate 3D body circumference measurements without relying on 2D approximations, as the 3D model captures the true geometric shape of the body cross-section
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention relates to a method for deriving accurate body size measures of a user from a sequence of 2D images, comprising the steps of: a) scanning the body of said user by obtaining a sequence of raw 2D images of said user as captured by a camera; b) analyzing the behavior of said user by tracking the position, pose/rotation and posture of said user before and during the scanning, by using a user behavior analyzer (UBA); c) extracting and encoding 2D shape data descriptors from said sequence of images by using a 2D shape analyzer (2DSA); and d) receiving said 2D shape descriptors as well as the user's position, pose and rotation data and integrating them into a 3D shape model, wherein said integration includes assigning rotation values to said sequence of images, wherein said values are either absolute values with respect to a full frontal position or relative, with respect to a reference image in said sequence.