Multi-Target Tracking Using Top-K Candidate Paths
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current sensor and data processing systems for detecting and tracking multiple objects undergoing simultaneous changes are bulky, costly, and power-intensive, and face challenges in determining individual object states and overall system state due to large data volumes and complex computations, especially when the number of objects is unknown.
Innovation Solution
The method involves generating candidate paths using a set of parameters to estimate the state of a system with multiple objects, where probabilities indicate the likelihood of paths and sensor observations corresponding to objects, with the number of candidate paths maximized based on processing and memory constraints, using techniques like Kalman filters and entropy maximization for efficient computation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the number of candidate paths is increased to improve tracking accuracy, then measurement precision is improved, but device complexity and computational load increase
Solution Approach 1:
The patent applies partial action by generating a limited number of candidate paths (e.g., top K paths) rather than exhaustively evaluating all possible paths. This selective approach maintains sufficient tracking accuracy while significantly reducing computational complexity, allowing the system to process multiple targets efficiently without requiring excessive computational resources.
Solution Approach 2:
The patent changes parameters by using probabilistic representations (probability distributions over candidate paths) and adjusting the number of candidate paths based on performance and memory constraints. This allows dynamic optimization of the trade-off between track accuracy and computational load by tuning parameters such as the number of candidate paths and probability thresholds.
2Measurement precision
If the number of candidate paths is increased to improve tracking accuracy, then measurement precision is improved, but memory consumption increases
Solution Approach 1:
The patent generates only the necessary number of candidate paths (top K) rather than maintaining all possible paths in memory. This partial action approach ensures that memory consumption remains manageable while still providing sufficient candidate paths to achieve accurate tracking, especially in multi-target scenarios where exhaustive enumeration would be infeasible.
Solution Approach 2:
The patent dynamically adjusts the number of candidate paths based on memory constraints and performance requirements. By changing this parameter, the system can optimize the balance between track accuracy and memory usage, allowing deployment on systems with limited memory resources while maintaining acceptable tracking performance.
3Measurement precision
If more sensor observations are processed to improve tracking accuracy, then measurement precision is improved, but productivity and processing time decrease
Solution Approach 1:
The patent segments the processing task by separately determining two sets of probabilities: (1) the likelihood that each candidate path corresponds to a true object trajectory, and (2) the likelihood that each sensor observation corresponds to a particular object. This segmentation allows independent optimization of each probability set and enables efficient processing of large numbers of observations without proportionally increasing computational burden.
Solution Approach 2:
The patent processes sensor observations selectively by focusing computational effort on the most likely candidate paths and observations. Rather than exhaustively processing all possible path-observation combinations, the system processes a manageable subset that provides sufficient tracking accuracy, thereby maintaining high processing speed even with large volumes of sensor data.
4Adaptability or versatility
If the system is designed to handle unknown number of objects, then adaptability is improved, but device complexity increases
Solution Approach 1:
The patent creates a universal tracking system that can handle any number of objects by using probabilistic candidate path representations. The same core algorithm works whether there is one object or multiple objects, as the system dynamically adjusts the number of candidate paths and their associated probabilities based on the actual number of targets present. This multi-functionality eliminates the need for separate processing logic for different object counts.
Solution Approach 2:
The patent employs dynamic probability distributions over candidate paths that can adapt to changing numbers of objects. As new observations arrive or objects enter/leave the scene, the system dynamically updates the probability assignments and candidate path set, allowing flexible adaptation to unknown and changing object counts without requiring pre-specified knowledge of the number of targets.
Data Source
AI summary
In an apparatus for determining the state of a system in which several system components undergo respective changes simultaneously, sensor measurements obtained from the components and candidate paths representing the individual states of the different components are analyzed. In this analysis, the system state is modeled in terms of likelihoods that certain paths correspond to true parameterized paths representing individual states of the system components and likelihoods that certain observations are associated with certain components. An optimization of the model provides accurate values of each type of likelihood, which then indicate the likely state of the system.