Target Tracking Using Deep Recurrent Networks to Reduce Error Accumulation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing target tracking technologies rely heavily on reference images, leading to inefficiencies in prediction precision and increased error accumulation, particularly in scenarios where the target object's position changes significantly over frames.
Innovation Solution
A target tracking method that obtains features from multiple reference images, determines initial predicted positions based on these features, and calculates a final position using a deep recurrent network framework, which reduces dependence on reference images and enhances prediction accuracy by reusing intermediate feature maps.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If target tracking relies heavily on reference images, then the tracking process can be simplified, but prediction precision deteriorates and error accumulation increases
Solution Approach 1:
The patent segments the tracking process into multiple independent prediction stages, each processing a subset of reference images separately before aggregation. This divides the complex task of using all reference images into manageable segments, reducing computational complexity while maintaining precision through multi-stage processing
Solution Approach 2:
The patent introduces a temporal dimension by processing reference images at different time intervals and combining predictions across multiple stages. This multi-dimensional approach transforms the single-step prediction into a sequential process, improving precision without proportionally increasing complexity
2Reliability
If multiple reference images are used to improve prediction accuracy, then tracking robustness improves, but computational time increases
Solution Approach 1:
The patent implements periodic processing where reference images are handled in periodic intervals rather than all at once. The multi-stage prediction process processes subsets of reference images in periodic cycles, reducing peak computational time while maintaining robustness through cumulative processing of all images
Solution Approach 2:
The patent performs preliminary processing of reference images by extracting features and organizing them before the main prediction stage. This preliminary action prepares data in advance, reducing the computational burden during the actual tracking prediction and allowing multiple reference images to be used efficiently
Data Source
AI summary
Target tracking methods and apparatuses, electronic devices, and storage media are provided. The method includes: obtaining features of a plurality of reference images of a target image; determining a plurality of initial predicted positions of a tracking target in the target image based on the features of the plurality of reference images; and determining a final position of the tracking target in the target image based on the plurality of initial predicted positions.