3D Human Pose Modeling via Articulated Skeleton Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video surveillance systems face challenges in accurately detecting and modeling human shapes from multi-view imagery, particularly in maintaining smooth surface deformation and accurately tracking human poses, due to computational demands and noise in silhouettes.
Innovation Solution
The method involves receiving video streams, detecting human objects, determining 3D hulls, generating initial pose hypotheses, refining these hypotheses using coarse 3D human models, and selecting an optimum model by projecting and comparing silhouettes, allowing for the detection of human attributes and accessories.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If 3D mesh surfaces are deformed to model human shape variations, then detailed 3D shape modeling is achieved, but computational complexity increases and surface smoothness deteriorates due to noise in silhouettes
Solution Approach 1:
The patent segments the human body into articulated body parts (head, torso, limbs) with joint constraints, allowing independent transformation of each segment while maintaining overall coherence. This segmentation enables localized shape adjustments without requiring global mesh deformation, reducing computational complexity while preserving modeling accuracy.
Solution Approach 2:
The patent introduces an intermediary articulated skeleton structure that mediates between the observed silhouette and the final 3D mesh model. The skeleton serves as a constraint framework that guides mesh deformation, preventing noise propagation and maintaining surface smoothness while achieving accurate shape representation.
2Measurement precision
If direct 3D mesh deformation is used to capture body shape variations, then detailed shape information is obtained, but surface smoothness deteriorates due to noisy silhouettes
Solution Approach 1:
The articulated skeleton acts as an intermediary structure that decouples the noisy silhouette observation from the smooth mesh deformation process. By constraining mesh transformations to follow skeleton-defined joint rotations and segment transformations, the system maintains surface smoothness while capturing accurate shape variations.
Solution Approach 2:
The patent transforms the shape representation from direct mesh vertex coordinates to skeletal parameters (joint positions, rotation angles, segment lengths). This parameterization changes the degrees of freedom from thousands of mesh vertices to a manageable set of skeletal parameters, enabling smooth transformations that respect anatomical constraints.
3Measurement precision
If multiple pose hypotheses are generated and refined, then pose estimation accuracy improves, but processing time increases
Solution Approach 1:
The system performs preliminary action by generating multiple initial pose hypotheses from silhouette matching before refinement. This preliminary set of hypotheses provides a constrained search space for subsequent optimization, reducing the computational burden of finding the optimal pose while maintaining accuracy through multiple candidate evaluation.
Solution Approach 2:
The patent implements feedback mechanisms where each pose hypothesis is evaluated against silhouette consistency and anatomical constraints, with results fed back to refine subsequent hypotheses. This iterative feedback process converges to accurate pose estimates while avoiding exhaustive search through all possible poses.
Data Source
AI summary
A video surveillance system, device and methods may accurately model the shape of a human object monitored by a video stream. 3D human models, such as a coarse 3D human model and a detailed 3D human model may be estimated by mapping individual body part components to a frame. For example, a coarse 3D human model may be obtained by mapping the cylindrical body parts to a plurality of skeleton pose estimates on a part by part basis. A detailed 3D human model may be estimated by mapping detailed human body parts to respective the cylindrical body parts of the coarse 3D human model on a part by part basis. The detailed 3D human model may be used to detect accessories of the human object being monitored, as well as overall dimensions, body part dimensions, age, and gender of the human object being monitored.