Multi-Target Tracking with Dependent Likelihoods for Occlusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current tracking techniques assume independence of individual target-measurement associations, which can lead to decreased accuracy in situations where this independence does not hold, such as in collision or occlusion conditions.
Innovation Solution
The implementation of a process for multi-target tracking with dependent likelihood structures, which involves receiving sensor data, obtaining candidate hypotheses, evaluating these hypotheses against criteria such as collision and occlusion structures, and determining a ranked set of hypotheses to estimate target locations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If independence assumption is used for target-measurement associations, then computational complexity is reduced, but tracking accuracy deteriorates in collision or occlusion conditions
Solution Approach 1:
The patent segments the hypothesis evaluation process into two stages: first evaluating each hypothesis independently for computational efficiency, then applying a verification stage that checks for collision and occlusion conditions. This segmentation allows the system to benefit from both independent evaluation speed and dependent evaluation accuracy where needed.
Solution Approach 2:
The patent performs preliminary independent evaluation of all candidate hypotheses before conducting the more computationally intensive verification step. By doing the quick independence check first, the system eliminates obviously poor hypotheses early, reducing the number of hypotheses that need full verification and thus lowering overall computational complexity while maintaining accuracy.
2Measurement precision
If all candidate hypotheses are evaluated without reduction, then tracking accuracy is maintained, but computational resources increase
Solution Approach 1:
The patent extracts and removes hypotheses that fail the verification criteria (collision or occlusion conditions) from the candidate set. By taking out these invalid hypotheses early in the process, the system reduces the number of hypotheses that need to be fully evaluated and tracked, improving computational efficiency while maintaining tracking accuracy for valid hypotheses.
Solution Approach 2:
The patent applies partial evaluation by performing independent likelihood assessment on all hypotheses first, then applying verification only where needed. This partial action approach avoids the excessive computation of fully evaluating every hypothesis with complete dependency checks, while still ensuring accuracy for the hypotheses that matter most.
3Productivity
If independent likelihood structures are used, then computational resources are conserved, but ability to detect collision and occlusion conditions deteriorates
Solution Approach 1:
The patent performs preliminary independent evaluation to identify promising hypotheses, then applies verification checks for collision and occlusion conditions on this reduced set. This preliminary action allows the system to conserve computational resources on obvious failures while ensuring reliable detection of collision and occlusion conditions for the hypotheses that warrant further investigation.
Solution Approach 2:
The verification stage acts as an intermediary between the independent likelihood evaluation and the final tracking decision. This intermediary layer checks for collision and occlusion conditions that the independent evaluation would miss, while the system overall remains computationally efficient because the intermediary only processes hypotheses that passed the initial independent evaluation threshold.
Data Source
AI summary
The present disclosure includes systems, methods, and computer-readable storage media facilitating multi-target tracking with dependent likelihood structures. To facilitate tracking, sensor data associated with an environment that a vehicle is located in may be received. A set of candidate hypotheses may be obtained based on the sensor data. Each candidate hypothesis in the set of candidate hypotheses may include track and measurement information representative of candidate locations for one or more targets. The set of candidate hypotheses are evaluated against at least one criterion. A ranked set of hypotheses are determined based on the evaluation of the set of candidate hypothesis against the at least one criterion. The ranked set of hypotheses may include a subset of hypotheses selected from among the set of candidate hypotheses and may be used to estimate a location of each of the one or more targets.