Articulated Object Tracking via Super-pixel Connectivity Graphs

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current object tracking technologies face challenges in accurately tracking articulated objects, such as humans, in video sequences due to their changing geometrical shape and motion states, leading to issues like 'fading-out' when parts of the object stop moving, as they rely solely on temporal derivatives without considering the object's complete structure.

Innovation Solution

An image-processing apparatus that segments each frame into super-pixels, constructs connectivity graphs, and uses temporal derivative maps to track articulated objects by associating non-zero derivatives with super-pixels, updating classifications based on previous frames, and restoring missing parts based on connectivity graph comparisons.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If temporal derivatives are used to track articulated objects, then moving objects can be detected, but objects fade out when parts stop moving

Engineering Contradiction:
Improveobject detection accuracyVSAvoidtracking continuity
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent segments the articulated object into multiple parts and tracks each part independently using connectivity graphs. This allows the system to maintain tracking of stationary parts even when temporal derivatives fail to detect them, preventing the fading-out problem while preserving accurate detection of moving parts.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary segmentation of the articulated object into parts before tracking begins. By pre-establishing the structural relationships between parts through connectivity graphs, the system can maintain continuous tracking of all parts regardless of their motion state, avoiding the reliability issue of temporal derivative-based methods.

Inventive Principle:
Principle #10Preliminary action

2Ease of manufacture

If background subtraction is used to locate foreground objects, then simple segmentation is achieved, but articulated objects with changing shapes cannot be accurately tracked

Engineering Contradiction:
Improvesegmentation simplicityVSAvoidobject boundary accuracy
Core Design Contradiction:
Ease of manufactureVSMeasurement precision

Solution Approach 1:

The patent divides the articulated object into multiple parts and uses connectivity graphs to represent spatial relationships between these parts. This segmentation approach maintains simple processing while accurately capturing the changing boundaries of articulated objects as they move and deform.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent employs dynamic connectivity graphs that are updated across video frames to adapt to changing object shapes and configurations. This allows the system to maintain accurate object boundaries even as articulated objects deform, while keeping the segmentation process computationally efficient.

Inventive Principle:
Principle #15Dynamics

3Ease of operation

If average movement from center of mass is used to track objects, then simple tracking is achieved, but articulated objects with changing geometrical shapes cannot be accurately tracked

Engineering Contradiction:
Improvetracking simplicityVSAvoidobject position accuracy
Core Design Contradiction:
Ease of operationVSMeasurement precision

Solution Approach 1:

The patent segments the articulated object into multiple parts and tracks each part's position and motion independently. This maintains operational simplicity while significantly improving position accuracy by accounting for the changing geometrical shape and differential motion of different body parts.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extends tracking from a single center-of-mass point to multiple dimensional components (individual parts) of the articulated object. This multi-dimensional approach preserves the simplicity of point-tracking operations while achieving accurate representation of complex articulated motion patterns.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Data Source

PatentEP3537377B1Automated tracking and retaining of an articulated object in a sequence of image frames
Publication Date: 2022.05.11 SONY GROUP CORP
  • EP3537377B1 patent drawingFigure 1A
  • EP3537377B1 patent drawingFigure 1B
  • EP3537377B1 patent drawingFigure 2

AI summary

Various aspects of an image-processing apparatus and method to track and retain an articulated object in a sequence of image frames are disclosed. The image-processing apparatus is configured to segment each image frame in the sequence of image frames into different segmented regions that corresponds to different super-pixels. An articulated object in a first motion state is detected by non-zero temporal derivatives between a first image frame and a second image frame. A first connectivity graph of a first set of super-pixels of the first image frame, is constructed. A second connectivity graph of a second set of super-pixels of the second image frame, is further constructed. A complete object mask of the articulated object in a second motion state is generated based on the first connectivity graph and the second connectivity graph, where at least a portion of the articulated object is stationary in the second motion state.