Subject Tracking Using Similarity Factor Difference for Search Area Adjustment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing subject tracking methods using pattern matching can result in false matching due to background similarities with the template, leading to inaccurate tracking.
Innovation Solution
A method that calculates a similarity factor between image frames and a template, determines the subject position based on the highest similarity, and adjusts the search target area by considering the difference between the highest and second-highest similarity factors, setting multiple search areas if the difference is low to improve tracking accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If template matching is executed to determine subject position, then subject tracking is achieved, but false matching occurs when background contains similar image portions
Solution Approach 1:
The patent divides the image into multiple search target areas and performs template matching independently in each area. By segmenting the image, the system can identify multiple potential matches and distinguish them based on their individual similarity factors, preventing false matching when the background contains regions similar to the template.
Solution Approach 2:
The patent calculates similarity factors for different search target areas and uses these local quality metrics to determine the most reliable subject position. Areas with higher similarity factors are prioritized, and the system can switch between areas based on real-time similarity assessments, ensuring accurate tracking even when background regions resemble the template.
2Device complexity
If a single search target area is used for template matching, then processing is simplified, but false matching occurs due to background similarities
Solution Approach 1:
The patent segments the search space into multiple search target areas, each subjected to independent template matching. This segmentation increases reliability by providing multiple candidate positions while maintaining manageable complexity through systematic evaluation of each area's similarity factor.
Solution Approach 2:
The patent dynamically adjusts the search process by evaluating similarity factors in real-time and switching between different search target areas based on the highest similarity values. This dynamic approach maintains high reliability while adapting to changing image content, preventing false matching without requiring overly complex static search designs.
3Productivity
If template matching is performed without considering multiple similarity factors, then processing is faster, but tracking accuracy decreases due to false matches
Solution Approach 1:
The patent performs template matching in segmented search target areas and calculates similarity factors for each, enabling parallel processing that maintains speed while improving precision through multi-factor evaluation of potential subject positions.
Solution Approach 2:
The patent uses feedback from similarity factor calculations to guide the search process, selecting areas with the highest similarity values as the most reliable subject positions. This feedback mechanism ensures high measurement precision by continuously evaluating and selecting the best matches without significantly slowing processing.
Data Source
AI summary
A subject tracking method, includes: calculating a similarity factor indicating a level of similarity between an image contained in a search frame at each search frame position and a template image by shifting the search frame within a search target area set in each of individual frames of input images input in time sequence; determining a position of the search frame for which a highest similarity factor value has been calculated, within each input image to be a position (subject position) at which a subject is present; tracking the subject position thus determined through the individual frames of input images; calculating a difference between a highest similarity factor value and a second highest similarity factor value; and setting the search target area for a next frame based upon the calculated difference.