Uncertainty-Aware Bounding Box Fusion for Stable Object Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current sensor fusion systems in autonomous vehicles face instability and poor tracking performance due to over- or under-estimation of sensor data uncertainty, leading to jittery or inaccurate fused bounding boxes.
Innovation Solution
The system generates a fused object bounding box based on uncertainty by comparing parameters from multiple sensor-based bounding boxes, using refined or consolidation fusion techniques to account for discrepancies, resulting in a fused bounding box with a confidence level of associated uncertainty.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If sensor fusion systems combine information from multiple sensors to improve tracking accuracy, then the reliability of object detection is improved, but the stability of fused bounding box estimation deteriorates due to over- or under-estimation of uncertainty
Solution Approach 1:
The patent changes the parameter representation from fixed bounding box values to probability distributions (Gaussian distributions) that capture uncertainty. By representing each sensor's bounding box as a distribution with mean and covariance, the system can properly quantify and fuse uncertainty, resolving the instability caused by over- or under-estimation of uncertainty in traditional fixed-value fusion approaches.
2Measurement precision
If the system uses refined fusion (intersection) to improve precision of object location, then measurement precision is improved, but reliability deteriorates when sensor discrepancies are large
Solution Approach 1:
The patent implements dynamic fusion strategy that adapts between refined fusion and consolidation fusion based on the observed discrepancy between sensors. When sensor agreements are high, refined fusion (intersection) is used for precision; when discrepancies are large, consolidation fusion (union) is used for reliability. This dynamic adaptation resolves the contradiction by selecting the appropriate fusion mode based on real-time sensor performance.
3Reliability
If the system uses consolidation fusion (union) to improve reliability of object detection, then tracking reliability is improved, but measurement precision deteriorates
Solution Approach 1:
The patent implements dynamic fusion strategy that adapts between refined fusion and consolidation fusion based on the observed discrepancy between sensors. When sensor agreements are high, refined fusion (intersection) is used for precision; when discrepancies are large, consolidation fusion (union) is used for reliability. This dynamic adaptation resolves the contradiction by selecting the appropriate fusion mode based on real-time sensor performance.
Data Source
AI summary
This document describes techniques and systems for generating a fused object bounding box based on uncertainty. At least two bounding boxes, each associated with a different sensor, is generated. A fused center point and yaw angle as well as length, width, and velocity can be found by mixing the distributions of the parameters from each bounding box. A discrepancy between the center points of each bounding box can be used to determine whether to refine the fused bounding box (e.g., find an intersection between at least two bounding boxes) or consolidate the fused bounding box (e.g., find a union between at least two bounding boxes). This results in the fused bounding box having a confidence level of the uncertainty associated with the fused bounding box. In this manner, better estimations of the uncertainty of the fused bounding box may be achieved to improve tracking performance of a sensor fusion system.