UAV Target Tracking via Detector and Tracker Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional target tracking techniques for UAVs face challenges such as loss of information due to 2D projection from 3D, noise, occlusions, and abrupt scene changes, especially in dynamic environments, limiting their robustness and adaptability.
Innovation Solution
A system and method that combines detector and tracker fusion using a detector and tracker module with a detector and tracker fusion (D&TF) module, incorporating feature encoding and attention mechanisms to generate robust target predictions, overcoming traditional tracking limitations by fusing classical and novel vision-based detectors and trackers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional tracking techniques are used, then device complexity is reduced, but tracking reliability deteriorates in dynamic environments with occlusions and scene changes
Solution Approach 1:
The patent combines detector and tracker modules into a unified system that processes visual information through multiple stages. The detector identifies candidate objects while the tracker maintains continuous tracking, and their fusion creates a more reliable tracking system that handles occlusions and scene changes better than conventional single-method approaches.
Solution Approach 2:
The tracking system is divided into distinct functional modules: detector module for initial object identification, tracker module for continuous tracking, and fusion module for integrating their outputs. This segmentation allows each module to specialize in specific tasks, improving overall reliability while managing complexity through modular design.
2Adaptability or versatility
If detector and tracker fusion is implemented, then tracking adaptability improves in unconstrained environments, but device complexity increases
Solution Approach 1:
The fusion module dynamically adjusts the contribution of detector and tracker outputs based on current scene conditions. During occlusions or abrupt scene changes, the system adapts by relying more on tracker predictions, while in stable conditions it incorporates more detector information, enabling versatile performance across different environmental constraints.
Solution Approach 2:
The unified detector-tracker fusion system serves multiple functions: initial object detection, continuous tracking, occlusion handling, and prediction of future object positions. This multi-functionality provides broad adaptability across various tracking scenarios without requiring separate specialized systems for each condition.
3Measurement precision
If robust tracking with feature encoding and attention mechanisms is used, then measurement precision improves, but use of energy increases
Solution Approach 1:
The system performs feature encoding and extracts relevant visual features in advance before the tracking decision process. By pre-processing and encoding features upfront, the attention mechanisms can operate more efficiently on already-processed data, reducing redundant computations and energy consumption during the main tracking loop while maintaining precision.
Data Source
AI summary
A system and a method for enhancing target tracking via detector and tracker fusion for unmanned aerial vehicles (UAVs) are provided. The method comprises receiving at least one raw input image of objects to be detected; based on the at least one raw input image of objects, generating the objects' candidate information; based on the objects' candidate information, calculating location and velocity estimation of an object at a current timestamp based on a detector and tracker fusion; and based on the location and velocity estimation of the object at the current timestamp, predicting the location and velocity estimation of the object at a future timestamp.