Multi-Camera Object Tracking via Pose-Specific Signature Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multi-camera object tracking systems face challenges in accurately matching object signatures across different camera views due to variations in pose and the need for a fixed number of exemplars, leading to errors and computational inefficiencies.
Innovation Solution
A method that determines correspondence between objects in different camera views by comparing spatial, temporal, and appearance characteristics, using a location-based signature selection approach to establish area relationships and match signatures from known poses, thereby reducing errors and computational costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the last-seen signature method is used to determine corresponding tracks, then the system requires only a single signature to store the representation of a track, but the method cannot compensate for differences in pose due to the position of the camera relative to the orientation of the tracked object, leading to errors in determining corresponding tracks
Solution Approach 1:
The patent segments the signature storage into multiple pose-specific signatures instead of using a single last-seen signature. Each pose-specific signature is stored separately according to the camera position and object orientation, allowing the system to select the appropriate signature based on the current pose, thereby maintaining accuracy while managing complexity through structured organization
Solution Approach 2:
The patent performs preliminary action by pre-capturing and storing multiple signatures of the tracked object from different poses and camera positions before the actual tracking matches are performed. This allows the system to have ready-to-use signatures for various orientations, eliminating the need to rely solely on the last-seen signature and improving match accuracy without requiring complex real-time pose estimation
2Reliability
If the exemplar method is used to determine corresponding tracks, then the system associates a predetermined set of signature exemplars with each track to represent different appearances, but the number of exemplars may have a fixed upper bound and the correct choice of the number of exemplars is critical, leading to computational expense when matching a given signature with each exemplar in turn
Solution Approach 1:
The patent applies preliminary action by pre-organizing signatures into pose-specific categories and pre-determining which signatures correspond to which poses. This eliminates the need to compare against all exemplars during matching, as the system can directly retrieve the appropriate pose-specific signature, significantly improving matching speed while maintaining representation accuracy
Solution Approach 2:
The patent applies local quality by creating pose-specific signature representations tailored to each camera position and object orientation. Instead of using a uniform set of exemplars for all poses, the system maintains specialized signatures for each local pose condition, improving matching accuracy for each specific situation without requiring a large universal exemplar set
3Adaptability or versatility
If multiple exemplars are used to represent different appearances of an object, then the system can better handle pose variations, but the number of exemplars increases computational cost and the fixed upper bound may not adequately represent all appearance variations
Solution Approach 1:
The patent applies dynamics by making the signature selection adaptive to the current pose rather than relying on a fixed number of static exemplars. The system dynamically selects the appropriate signature based on the detected pose and camera position, allowing it to handle any pose variation without being constrained by a predetermined upper bound on the number of exemplars
Solution Approach 2:
The patent changes the parameter of signature selection from a fixed exemplar index to a pose-dependent parameter. By using camera position and object orientation as parameters to select the appropriate signature, the system achieves high adaptability to pose variations while simplifying management, as the signature selection becomes a straightforward parameter lookup rather than a complex exemplar selection process
Data Source
AI summary
Disclosed herein are a method, system, and computer program product for determining a correspondence between a first object (713) tracked in a first field of view and a second object tracked (753) in a second field of view. The method determines a first area (711) in the first field of view, based on the location and size of the first object (713). The method utilizes a predetermined area relationship between the first area (711) in the first field of view and at least one area (751) in the second field of view to determine a second area (751) in the second field of view. In one embodiment, the method determines the second area (751) in the second field of view by comparing predetermined area relationships between the first area (711) and any areas (751) in the second field to determine a best match. The method determines a correspondence between the first object (713) and the second object (753), based on a comparison between a first object signature associated with the first object (713) and a second object signature associated with the second object (753).