Object Tracking Template Verification for Appearance Changes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object tracking systems face challenges in maintaining tracking performance due to changes in the appearance of the target object, often leading to inappropriate template updates that degrade tracking accuracy.
Innovation Solution
The method employs a combination of forward and backward object tracking using a shared object tracking model to evaluate template candidates based on confidence scores derived from overlap and historical tracking results, ensuring accurate template updates by verifying suitability through backward tracking.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If template updates are performed frequently to adapt to appearance changes, then adaptability improves, but tracking accuracy deteriorates due to inappropriate template updates
Solution Approach 1:
The patent performs backward object tracking in advance before final template update decision. This preliminary verification step checks whether a template candidate would be appropriate by tracking backwards from the current frame using the candidate template. Only if the backward tracking results are consistent with forward tracking results is the template update performed, thus preventing inappropriate updates while maintaining adaptability to legitimate appearance changes.
2Reliability
If forward object tracking is performed continuously to maintain tracking results, then tracking performance is maintained, but computational resources are consumed
Solution Approach 1:
The patent implements periodic template updates rather than continuous updates. Template updates are performed at specific intervals or under specific conditions (when appearance changes are detected and verified through backward tracking). This periodic approach maintains tracking performance when needed while reducing computational resource consumption during stable periods.
3Measurement precision
If backward object tracking is performed to verify template candidates, then template update accuracy improves, but processing time increases
Solution Approach 1:
The patent uses backward tracking results as an intermediary verification mechanism. Instead of directly accepting template candidates based on forward tracking alone, the system introduces backward tracking as an intermediate validation step. This intermediary process checks consistency by tracking backwards from the current frame, ensuring template update accuracy while minimizing processing time through efficient consistency checking rather than full re-tracking.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A processor-implemented method for object tracking includes: performing, using a first template, forward object tracking on first image frames in a first sequence group; determining a template candidate of a second template for second image frames in a second sequence group; performing backward object tracking on the first image frames using the template candidate; determining a confidence of the template candidate using a result of comparing a first tracking result determined by the forward object tracking performed on the first image frames and a second tracking result determined by the backward object tracking performed on the first image frames; determining the second template based on the confidence of the template candidate; and performing forward object tracking on the second image frames using the second template.