Shadow-Assisted Object Tracking for Occluded Video

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Video object tracking is complex due to fast-moving objects, occlusions, and changes in orientation, which complicates the detection and association of objects across video frames, especially in scenarios like sporting events where occlusions and multiple players obscure each other.

Innovation Solution

The method utilizes shadow information from objects to improve object recognition and tracking by segmenting, localizing, and extracting 3D information, including the use of dominant shadow directions to accurately determine object positions and poses, even in occluded conditions, by analyzing shadow projections and their intersections to refine object location estimates.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional background subtraction and feature matching techniques are used for object tracking, then the system can operate with standard video processing, but tracking accuracy deteriorates in occluded scenarios and when objects move fast relative to frame rate

Engineering Contradiction:
Improvetracking accuracyVSAvoidsystem complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent segments the object into multiple parts by tracking extreme points (corners, tips, ends) rather than treating the object as a whole. This segmentation allows each extreme point to be tracked independently, improving accuracy during occlusions and fast motion while maintaining manageable system complexity through focused feature tracking.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces shadow information as an intermediary element to assist object tracking. By detecting and analyzing shadows cast by objects, the system gains additional spatial and temporal information that helps maintain tracking accuracy during occlusions and fast motion, without requiring complex multi-camera systems.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If object recognition techniques are added to video tracking to handle fast motion and orientation changes, then tracking robustness improves, but processing time increases due to the complexity of analyzing multiple features across frames

Engineering Contradiction:
Improvetracking robustnessVSAvoidprocessing time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent applies partial action by tracking only extreme points rather than all features of an object. This selective approach provides sufficient tracking robustness for fast motion and orientation changes while significantly reducing processing time compared to full object recognition techniques.

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The system performs preliminary detection of extreme points in each frame before attempting to associate them with previous frames. This preliminary identification simplifies the subsequent association process and reduces overall processing time while maintaining tracking robustness.

Inventive Principle:
Principle #10Preliminary action

3Productivity

If multiple objects are tracked simultaneously, then the system provides comprehensive scene monitoring, but association difficulty increases when objects are close together or occluding each other

Engineering Contradiction:
Improvescene monitoring coverageVSAvoidobject association difficulty
Core Design Contradiction:
ProductivityVSDifficulty of detecting and measuring

Solution Approach 1:

The patent segments the tracking problem into independent extreme point tracking tasks. By treating each extreme point as a separate tracking entity, the system can monitor multiple objects simultaneously while simplifying the association process, as extreme points provide distinctive features that are easier to match even when objects are close or occluded.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Shadow detection serves as an intermediary that provides additional spatial information to resolve ambiguities in object association. When multiple objects are present and occluding each other, shadow information helps distinguish between them and correctly associate extreme points with their source objects.

Inventive Principle:
Principle #24Intermediary (Mediator)

4Adaptability or versatility

If the tracked object changes orientation and pose over time, then the tracking system can handle more dynamic scenarios, but the complexity of maintaining accurate object models increases

Engineering Contradiction:
Improvehandling dynamic scenariosVSAvoidobject model complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent applies partial action by focusing tracking on extreme points rather than maintaining complete object models. This approach handles dynamic scenarios with changing orientation and pose effectively while avoiding the complexity of full object modeling, as extreme points provide sufficient information for tracking without requiring detailed pose estimation.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP4358035A1Apparatus, systems and methods for shadow assisted object recognition and tracking
Publication Date: 2024.04.24 DISNEY ENTERPRISES INC
  • EP4358035A1 patent drawingFigure 1
  • EP4358035A1 patent drawingFigure 2
  • EP4358035A1 patent drawingFigure 3

AI summary

Described herein are apparatus, systems and methods for shadow assisted object recognition and tracking. The methods performed by the apparatus and system include identifying a blob within a video image, the video image having at least one object and at least one shadow of the at least one object, the at least one shadow of the at least one object cast by at least one light source. Identifying the blob includes identifying an object projection corresponding to the at least one object and a shadow projection corresponding to the at least one shadow. A location of an object portion of the at least one object is determined based on the shadow projection.