Object Tracking with FEM State-Space for Adaptive 3D Resolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current object tracking methods, particularly in Bayesian tracking schemes, face challenges such as large errors due to Gaussian modeling assumptions, difficulties in handling large areas of low probability density, and struggles with dynamic changes in state-space resolution and 3D handling.
Innovation Solution
The method employs a finite element model (FEM) representation to partition the state-space of interest, allowing for a high degree of freedom in modeling and enabling accurate state-space modeling. It updates the state-space distribution based on sensor and external data, and propagates this distribution over time, while also adjusting the state-space resolution as needed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If Gaussian modeling is used in Kalman filters, then the tracking computation is simplified, but large modeling errors are introduced due to incorrect distribution assumptions
Solution Approach 1:
The patent changes the fundamental parameter of probability distribution from Gaussian to a more flexible representation that can adapt to the true distribution shape. This is achieved by using a state-space model with probability density functions that are not constrained to Gaussian forms, thereby reducing modeling errors while maintaining computational tractability through the structured state-space framework.
2Measurement precision
If particle filters are used to reduce modeling errors, then distribution accuracy improves, but difficulties arise in handling large areas of low probability density
Solution Approach 1:
The patent segments the state-space into discrete states with associated probability densities, allowing selective refinement in regions of interest. This segmentation enables the model to focus computational resources on areas with significant probability mass while efficiently representing low-density regions through coarser discretization or analytical solutions.
3Measurement precision
If particle methods are used for tracking, then modeling errors are reduced, but dynamic changes in state-space resolution cannot be applied due to lack of built-in interconnection between nodes
Solution Approach 1:
The patent implements a dynamic state-space representation where the resolution and discretization can be adapted over time based on the tracking needs and probability distribution characteristics. The model allows dynamic refinement of state-space resolution in regions of high probability density while maintaining coarser representation elsewhere, enabling flexible adaptation to changing tracking conditions.
4Ease of manufacture
If traditional models are used for tracking, then computation is simpler, but accurate definition of 3D state-spaces such as altitude maps is difficult
Solution Approach 1:
The patent extends the state-space model to explicitly handle three-dimensional spaces including altitude or elevation dimensions. By incorporating vertical dimension into the state-space definition with appropriate boundary conditions and probability density functions, the model can accurately represent 3D tracking scenarios such as altitude maps while maintaining the structured computational framework.
Data Source
AI summary
The present disclosure relates to a computer-implemented method for object tracking, the method including the steps of defining a state-space of interest based on a class of objects subject to tracking. Further, the method includes the step of representing the state-space of interest using a FEM representation partitioning the state-space of interest in elements. Further, the method includes initiating a state-space distribution defining a probability density for different states of at least one tracked object in the state-space of interest. Moreover, the method updates the state-space distribution based on evidence, wherein the evidence being at least one of sensor data and external data of at least one tracked object in said class of objects. Furthermore, the method propagates the state-space distribution of the at least one tracked object for a time period.