3D Multi-View Object Tracking With Occlusion Error Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object tracking technologies fail to accurately track objects in crowded or occluded conditions, such as in ball sports, leading to errors and erroneous tracking results due to players or structures obstructing the view.
Innovation Solution
An information processing apparatus that utilizes multiple image capturing devices to obtain multi-viewpoint frames, estimates a three-dimensional shape of the object, tracks the object using identifiers, and employs identification processing to correct tracking errors by comparing results from tracking and identification units, resetting and complementing tracking data when errors are detected.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If tracking processing is used to continuously estimate object position, then tracking speed and continuity are improved, but accuracy deteriorates when objects are occluded or crowded
Solution Approach 1:
The system performs identification processing at specific intervals to verify tracking results, creating a feedback mechanism that detects and corrects tracking errors. The detection unit compares identification results with tracking results to determine if correction is needed, ensuring accuracy while maintaining continuous tracking operation.
Solution Approach 2:
The system executes identification processing periodically rather than continuously, balancing computational load with accuracy requirements. This periodic verification corrects accumulated tracking errors without sacrificing tracking continuity, resolving the contradiction between speed and precision.
2Measurement precision
If identification processing is performed frequently to correct tracking errors, then tracking accuracy is improved, but computational load increases
Solution Approach 1:
The detection unit intelligently determines when identification processing is needed by comparing tracking results with identification results, executing identification only when errors are detected. This feedback-based approach maintains accuracy while minimizing unnecessary computational expenditure.
Solution Approach 2:
The system dynamically adjusts the frequency of identification processing based on tracking confidence and detected errors, changing the processing parameter (identification frequency) to optimize the balance between accuracy and computational load for different operational conditions.
3Reliability
If multiple image capturing devices are used to capture objects from different directions, then tracking reliability in crowded scenarios is improved, but device complexity increases
Solution Approach 1:
The system transitions from two-dimensional image analysis to three-dimensional shape estimation by integrating data from multiple viewpoints. This dimensional transformation enables more reliable object identification and tracking in crowded scenarios, as the 3D shape remains distinctive even when 2D appearances are occluded or similar.
Solution Approach 2:
The multi-viewpoint image capturing system serves multiple functions: initial object identification, continuous tracking support, and error detection. This multi-functionality justifies the increased device complexity by providing comprehensive tracking reliability across different operational conditions.
Data Source
AI summary
An information processing apparatus according to the present disclosure obtains a plurality of captured images obtained by capturing images of an object from different directions, estimates a three-dimensional shape of the object using the plurality of captured images, performs tracking processing of the object to track the object by estimating an identifier and a position of the object using the three-dimensional shape, performs identification processing of the object to identify the object using at least some of the plurality of captured images and estimate an identifier and a position of the object, detects an error in the result of the tracking processing based on the result of the tracking processing and the result of the identification processing, and complements tracking data indicating the result of the tracking processing using the result of the identification processing in a case where an error in the tracking processing is detected.