Multiple Hypothesis Tracking via Maximum Weight Independent Set
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The reconstitution of global hypotheses in track-based multiple hypothesis tracking (MHT) is time-consuming and inefficient, particularly when dealing with clusters containing singleton tracks, leading to potential loss of important target tracks due to the NP-hard nature of the problem and the lack of performance guarantees in existing ad-hoc approaches.
Innovation Solution
Formulating the problem as an instance of the maximum weight independent set problem (MWISP) and using a system that receives tracking data, generates a matrix array, and performs local search moves to increase the global hypothesis score, allowing for efficient grouping and identification of target objects with reduced processing time and memory usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional ad-hoc approaches (breadth-first search, depth-first search) are used for reconstitution of global hypotheses, then the implementation is simple, but the processing time is excessive and performance is not guaranteed
Solution Approach 1:
The patent transforms the reconstitution problem from a traditional search problem into a Maximum Weight Independent Set Problem (MWISP) by changing the parameter representation. Tracks are represented as vertices in a graph with weights corresponding to hypothesis scores, and compatibility relationships define edges. This parameter transformation enables the use of efficient graph-based algorithms that can quickly identify high-ranking global hypotheses without exhaustive search.
Solution Approach 2:
The patent replaces the mechanical search-based approaches (breadth-first search, depth-first search) with a graph-theoretic formulation. Instead of systematically exploring the hypothesis space through search operations, the system constructs a compatibility graph where the solution emerges from identifying the maximum weight independent set, substituting the mechanical search process with a mathematical optimization approach.
2Reliability
If track-oriented approach with reconstitution of global hypotheses is used, then tracking accuracy can be improved, but the computational complexity increases significantly
Solution Approach 1:
The patent changes the computational parameters by formulating the problem as MWISP, where the objective function directly incorporates hypothesis scores as vertex weights. This parameter transformation allows the system to prioritize high-scoring hypotheses naturally through the graph structure, achieving accurate tracking without the need for complex exhaustive search algorithms.
Solution Approach 2:
The patent segments the complex reconstitution problem into two manageable components: (1) constructing the compatibility graph where vertices represent tracks and edges represent compatibility relationships, and (2) solving the MWISP on this graph to identify the best global hypothesis. This segmentation reduces computational complexity by breaking down the NP-hard problem into graph construction followed by independent set identification.
3Measurement precision
If exhaustive search methods are used to find high ranking global hypotheses, then accuracy is improved, but processing time increases exponentially
Solution Approach 1:
The patent transforms the search objective from finding any valid hypothesis to finding the maximum weight independent set, where weights correspond to hypothesis scores. This parameter change enables the algorithm to directly optimize for high-ranking hypotheses through the graph formulation, achieving accurate ranking without exponential search time by leveraging the mathematical properties of the MWISP.
Data Source
AI summary
The technology described herein includes a system and/or a method for multiple hypothesis tracking. In some examples, a multiple hypothesis tracking system includes a receiver, an array generation module, and a local search move module. The receiver is configured to receive tracking data associated with one or more target families from one or more sensors. The array generation module is configured to generate a matrix array based on the tracking data. The local search move module is configured to perform one or more local search moves on one or more cells in the matrix array to increase a global hypothesis score for the tracking data.