LiDAR Bounding Box Heading Accuracy via Side Mirror Point Removal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle control systems using LiDAR to identify external objects often incorrectly determine the heading direction of external vehicles due to the presence of side mirrors, leading to potential errors in vehicle route changes or speed adjustments.
Innovation Solution
A vehicle control apparatus and method that generates a bounding box by excluding the portion corresponding to the side mirror of an external vehicle, allowing for accurate identification of the heading direction and type of the external vehicle.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If a virtual box is generated to include the side mirror of an external vehicle, then the bounding box covers the entire external vehicle, but the heading direction is incorrectly identified due to the side mirror
Solution Approach 1:
The patent extracts and removes points corresponding to the side mirror from the point cloud data before generating the bounding box. This is achieved by identifying points within a specific height range that corresponds to the side mirror's position and excluding them from the bounding box calculation, thereby eliminating the source of heading direction errors while maintaining comprehensive vehicle coverage.
Solution Approach 2:
The patent segments the point cloud data into different regions based on height, separating the side mirror region from the main vehicle body. By applying different processing rules to different segments (removing side mirror points, keeping vehicle body points), the system achieves both complete vehicle coverage and accurate heading direction identification.
2Measurement precision
If points with specific height are removed from the point cloud to exclude the side mirror, then the heading direction identification is improved, but the complexity of point cloud processing increases
Solution Approach 1:
The patent applies local quality by using height-specific filtering that targets only the side mirror region while preserving the rest of the vehicle points. This localized approach simplifies processing compared to global processing methods, as it only requires filtering within a specific height range rather than analyzing the entire point cloud structure.
3Measurement precision
If the bounding box is corrected by comparing angles and distances, then the accuracy of external vehicle identification is improved, but the computational time and processing complexity increase
Solution Approach 1:
The patent performs preliminary action by pre-calculating and storing the height range corresponding to the side mirror before actual bounding box generation. This preprocessing step allows the main algorithm to quickly filter side mirror points using simple height comparisons rather than complex geometric analysis, significantly reducing real-time processing time while maintaining high accuracy.
Solution Approach 2:
The patent uses simple geometric comparisons (angle and distance thresholds) as disposable verification steps. These lightweight checks quickly validate whether bounding box correction is needed without requiring complex or time-consuming analysis, enabling fast accuracy verification.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The solution accurately identifies the heading direction and type of external vehicles, reducing errors in vehicle control systems and improving the reliability of driving route changes and speed adjustments.
Implementation Method 1
obtain a point cloud corresponding to an external vehicle through a light detection and ranging (LiDAR)
Data Source
AI summary
A vehicle control apparatus may include a light detection and ranging device (LiDAR) and a processor. According to a method, the processor may obtain a plurality of LiDAR data points by removing points with a specific height or higher from a point cloud corresponding to an external vehicle, may generate a bounding box based on contour points, which indicate a periphery of the external vehicle, and may obtain a final bounding box obtained by correcting the bounding box based on an angle between a first heading direction of a virtual box and a second heading direction of the bounding box exceeding a reference angle, and based on a distance between a first location of the virtual box and a second location of the bounding box being smaller than or equal to a first reference distance.