Autonomous Vehicle Object Matching Using Superposition Metrics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current matching algorithms for autonomous driving systems, such as the Intersection over Union (IOU) method, are computationally intensive, require specialized software, and fail to distinguish between objects with equivalent or non-intersecting areas, leading to inefficiencies in processing large datasets and determining accurate object matches.
Innovation Solution
A superposition of metrics (SOM) method that combines volume match (VM), distance match (DM), and angle match (AM) using a multiplication of their respective calculations, based on the geometric parameters of bounding boxes, to determine the degree of matching between reference and test objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the Intersection over Union (IOU) method is used for matching objects, then the matching accuracy is improved, but the computational complexity and processing time increase significantly
Solution Approach 1:
The patent segments the matching process into three independent components: volume match (comparing bounding box volumes), distance match (comparing centers and orientations), and angle match (comparing yaw angles). Each component is calculated separately using simple geometric formulas, avoiding the computationally intensive intersection area calculations of traditional IOU methods while maintaining matching accuracy.
2Measurement precision
If the Intersection over Union (IOU) method is used for matching objects, then the matching accuracy is improved, but the processing speed decreases
Solution Approach 1:
The patent extracts only the essential geometric features (volume, center position, orientation, and yaw angle) needed for matching, discarding the computationally expensive intersection area calculation. By taking out only the necessary components and using simple multiplication to combine them, the system achieves fast processing speed while maintaining matching accuracy.
3Measurement precision
If traditional matching algorithms are used, then comprehensive object comparison is achieved, but specialized software libraries are required
Solution Approach 1:
The patent implements a self-contained matching system that calculates all necessary geometric parameters (volume, distance, angle) using basic mathematical operations available in standard programming languages. The system does not depend on specialized computer vision libraries or complex algorithms, making it easily implementable with standard software tools while maintaining comprehensive object comparison capabilities.
Data Source
AI summary
A method for matching a reference object and a test object includes providing a test object in a field of sensing of at least one sensor disposed at a vehicle. A volume match is determined based on a volume of the reference object and a volume of the test object. A distance match is determined based on a center and orientation of the reference object and a center and orientation of the test object. An angle match is determined based on a yaw angle of the reference object and a yaw angle of the test object. A superposition of the volume match, the distance match, and the angle match is determined based on a multiplication of the volume match, the distance match, and the angle match. A degree of matching of the reference object and the test object is determined based on the superposition.