Gaussian Object Tracking via Rotated Bounding Box
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object tracking algorithms struggle with accurately tracking non-rigid objects that rotate or transform, particularly in fast-paced battlefield environments, due to their reliance on upright rectangular models that do not account for orientation changes, leading to reduced precision and increased computational burden.
Innovation Solution
A deep-learning based object tracking method using a Gaussian distribution model to define the area of interest, which includes parameters for position, shape, and rotation angle, minimizing a loss function based on Kullback-Leibler divergence to refine predictions and reduce computational requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pixel-level prediction with object segmentation module is introduced to increase tracking accuracy, then measurement precision is improved, but device complexity and computational burden increase
Solution Approach 1:
The patent extracts only the essential parameters (center coordinates, major/minor axes, rotation angle) needed for rotated bounding box representation, eliminating the need for complex pixel-level segmentation while maintaining tracking accuracy. This reduces computational burden by focusing on key geometric parameters rather than processing entire pixel masks.
Solution Approach 2:
The patent performs preliminary action by directly predicting rotated bounding box parameters through the neural network without requiring post-processing steps like finding rotated bounding boxes from binary masks. This eliminates computationally intensive post-processing operations while maintaining prediction accuracy.
2Device complexity
If upright rectangular models are used for object tracking, then device complexity is reduced, but measurement precision deteriorates when objects rotate or transform
Solution Approach 1:
The patent applies dynamics by transitioning from static upright rectangular models to dynamic rotated bounding box models that can adapt to object orientation changes. The neural network predicts rotation angles and adjusted bounding box parameters in real-time, allowing the tracking model to dynamically follow objects regardless of their rotation or transformation state.
Solution Approach 2:
The patent changes the parameters used for bounding box representation from simple upright coordinates to rotated parameters including center coordinates, major/minor axes lengths, and rotation angles. This parameter transformation enables accurate tracking of rotating objects while maintaining model efficiency through direct parameter prediction.
3Productivity
If existing object tracking algorithms are applied to fast-paced battlefield environments, then productivity is maintained, but measurement precision deteriorates due to object rotation and transformation
Solution Approach 1:
The patent replaces traditional mechanical bounding box adjustment methods with a neural network-based system that directly predicts rotated bounding box parameters. This substitution enables faster processing by eliminating iterative adjustment steps while improving precision through learned parameter predictions that account for object rotation and transformation in real-time.
Data Source
AI summary
Devices, systems, methods, and instructions for object tracking based on deep-learning are provided, including pre-training a model for object tracking based on pre-input learning data, receiving a target image of which at least one area contains an image corresponding an object for tracking and a search image of which at least one area contains an image corresponding the object for tracking, and obtaining information on area for tracking regarding to the area corresponding to the object for tracking in the search image by applying the model for object tracking, wherein the area corresponding to the object for tracking is defined by a Gaussian distribution model, and the information on area for tracking includes parameter values of a plurality of parameters based Gaussian distribution corresponding to the area corresponding to the object for tracking.