Video Object Detection Using Dynamic Template Sets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video object detection and tracking methods face challenges in accurately tracking objects due to changes in lighting, scale, background interference, and occlusions, which reduce the accuracy of object identification and tracking.
Innovation Solution
The method involves detecting an object from a frame image of a video based on a target template set, which includes initial and subsequent frame information, and an interference template set to account for interfering objects. The method integrates image features to determine candidate areas and updates the template sets based on similarity thresholds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a single template is used for object tracking, then the tracking process is simple, but the accuracy decreases when objects undergo changes in lighting, scale, background interference, or occlusions
Solution Approach 1:
The patent divides the single template into multiple sub-templates, where each sub-template corresponds to a specific region or feature of the target object. This segmentation allows the system to handle different appearances of the object under varying conditions (lighting, scale, occlusion) by selecting appropriate sub-templates for matching, thereby improving tracking accuracy without requiring a completely different template management system
Solution Approach 2:
The patent implements dynamic template selection where the system automatically chooses which sub-templates to use based on current video frame characteristics. The template set is dynamically updated by adding new sub-templates when the object's appearance changes significantly, and by reweighting existing sub-templates based on their matching performance, allowing the system to adapt to changing conditions while maintaining manageable complexity
2Reliability
If multiple templates are used to improve accuracy, then the system can handle object variations, but the computational complexity and processing time increase
Solution Approach 1:
The patent applies local quality by creating sub-templates that focus on specific local features or regions of the object rather than using the entire object as a single template. This allows the system to perform matching operations on smaller, more distinctive local regions, reducing the computational burden while maintaining high accuracy in identifying the object even when overall appearance varies
Solution Approach 2:
The patent uses partial action by selectively applying only the necessary sub-templates for each tracking scenario rather than processing all possible templates. The system determines which sub-templates are most relevant based on current visual conditions and focuses computational resources on those, avoiding the excessive processing that would result from evaluating every possible template combination
3Reliability
If the template set is continuously updated to reflect object changes, then tracking accuracy improves, but the system becomes more sensitive to interference from similar objects
Solution Approach 1:
The patent extracts and separates the object of interest from similar interfering objects by creating distinct sub-templates for each. When similar objects appear in the video, the system can identify which sub-templates correspond to the target object and which correspond to interfering objects, allowing it to focus matching operations only on the correct sub-templates and ignore similar but incorrect matches
Solution Approach 2:
The patent converts the potential harm of similar objects into a benefit by using the appearance variations of similar objects as training data to create and refine sub-templates. By deliberately incorporating examples of similar objects into the template generation process and using interference templates to teach the system what to reject, the system learns to distinguish between the target object and similar objects, turning the interference problem into an opportunity to improve discrimination accuracy
Data Source
AI summary
A target object detection method and apparatus are provided. The target object detection method and apparatus are applicable to fields such as artificial intelligence, object tracking, object detection, and image processing. An object is detected from a frame image of a video including a plurality of frame images based on a target template set including one or more target templates.