Target Tracking Neural Network with Segmented Template Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing target tracking technologies face challenges in accurately distinguishing between targets and backgrounds, especially when the target has low objectness or the background has high objectness, leading to potential target misidentification and reduced tracking accuracy.
Innovation Solution
A target tracking method and apparatus utilizing a neural network to extract feature information, calculate similarity matrices, and estimate target position by applying templates to separate target and background information, with a backbone network trained for object classification to suppress background influence and emphasize target features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional target tracking methods are used, then the tracking process is simple, but the tracking accuracy deteriorates when targets have low objectness or backgrounds have high objectness
Solution Approach 1:
The patent segments the template matching process into multiple components: basic template matching, positive template matching (for target features), and negative template matching (for background suppression). This segmentation allows the system to separately process and weigh different feature types, improving accuracy by emphasizing target features while suppressing background interference.
Solution Approach 2:
The patent applies different quality weights to different regions and features within the template matching process. By assigning higher weights to positive template matches (target features) and lower or negative weights to negative template matches (background features), the system creates a differentiated quality assessment that improves tracking accuracy in challenging scenarios.
2Loss of information
If background information is included in template matching, then more information is available for matching, but background influence increases causing target misidentification
Solution Approach 1:
The patent extracts and separates background information from target information through negative template matching. By explicitly identifying and extracting background features, the system can then suppress their influence in the final scoring, preventing background dominance while retaining the benefit of complete information for analysis.
Solution Approach 2:
The patent dynamically adjusts the weighting parameters in the scoring function based on the relative importance of positive versus negative template matches. By changing the weight parameters, the system can emphasize target features over background features, effectively controlling the influence of background information on the final tracking decision.
Data Source
AI summary
A target tracking method and apparatus is provided. The target tracking apparatus includes a memory configured to store a neural network, and a processor configured to extract feature information of each of a target included in a target region in a first input image, a background included in the target region, and a searching region in a second input image, using the neural network, obtain similarity information of the target and the searching region and similarity information of the background and the searching region based on the extracted feature information, obtain a score matrix including activated feature values based on the obtained similarity information, and estimate a position of the target in the searching region from the score matrix.