Photographic Subject Tracking with Reliability-Based Search Adjustment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing photographic subject tracking methods using template matching can result in false matches when the background contains similar images, leading to inaccurate subject tracking.
Innovation Solution
A method that calculates similarity levels between input images and a template image, determines the subject position based on these levels, and adjusts the search area and calculation methods if the reliability of the similarity level is low, to improve tracking accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If template matching is used to determine photographic subject position, then tracking can be performed from frame to frame, but false matches occur when the background contains similar images, reducing tracking accuracy
Solution Approach 1:
The patent implements a reliability determination mechanism that analyzes the distribution of similarity levels across multiple candidate positions. When the difference between the highest similarity level and the average similarity level exceeds a threshold, the system identifies this as a reliable match. This feedback loop continuously monitors match quality and adjusts tracking decisions accordingly, preventing false matches while maintaining high tracking speed.
Solution Approach 2:
The patent dynamically changes the matching parameters by adjusting the threshold for accepting a match based on the reliability assessment. When high reliability is detected, the system accepts matches more readily; when reliability is low, it rejects matches even with high similarity scores. This parameter adaptation allows the system to maintain both speed and accuracy under varying conditions.
2Reliability
If the search area is expanded to cover the entire image, then false matches in the background can be detected, but the calculation time increases significantly
Solution Approach 1:
The patent divides the image into multiple candidate positions and calculates similarity levels independently for each position. This segmentation allows parallel processing of multiple locations simultaneously, reducing overall calculation time while still examining the entire image for potential matches.
Solution Approach 2:
The patent calculates similarity levels at multiple candidate positions beyond just the single highest match, examining a sufficient number of alternatives to determine reliability. This partial action approach checks enough positions to ensure reliability without performing exhaustive searches at every possible location, balancing thoroughness with efficiency.
3Measurement precision
If multiple candidate positions are evaluated to determine reliability, then false matches can be reduced, but the computational complexity increases
Solution Approach 1:
The patent replaces complex mechanical or manual analysis with automated computational methods. The reliability determination is achieved through algorithmic comparison of similarity levels and automatic threshold-based decision making, substituting what would otherwise require complex manual evaluation of multiple candidate positions.
Data Source
AI summary
A photographic subject tracking method comprises calculating similarity levels of images input sequentially to a template image, determining a photographic subject position in each of the input images based upon the similarity levels, tracking the photographic subject position through the input images, judging whether reliability of the similarity level calculated for the photographic subject position is high or low, and changing ways of at least one of calculating the similarity levels, determining the photographic subject position and tracking the photographic subject position, if the reliability of the similarity level is judged to be low.