Obstacle Velocity Determination Using Point Cloud Registration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current unmanned driving systems face inaccuracies in obstacle velocity calculation due to fluctuations in centroid position data from point cloud coordinates, leading to large errors in observed velocity.
Innovation Solution
A method for determining obstacle velocity by registering and matching point cloud data from two time points, using point cloud distribution information to improve registration efficiency and accuracy, even in incomplete data scenarios, by projecting data onto two-dimensional planes and calculating matching degrees to find optimal displacement expectations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If obstacle velocity is calculated based on centroid position changes from point cloud coordinates, then the calculation process is simple, but the measurement precision is poor due to large fluctuations and blocking issues
Solution Approach 1:
The patent segments the point cloud data into multiple clusters representing different obstacles, and calculates velocity for each cluster separately using their respective centroid positions. This segmentation approach allows the system to handle complex scenes with multiple objects, reducing interference between obstacles and improving overall measurement precision while maintaining computational efficiency
Solution Approach 2:
The patent introduces an intermediary matching process that aligns point cloud data from consecutive frames before calculating velocity. By using feature matching and transformation matrices as intermediaries to register points across time, the system eliminates fluctuations caused by blocking and improves velocity measurement accuracy without significantly increasing calculation complexity
2Adaptability or versatility
If point cloud data is used for obstacle tracking, then the system can handle complex 3D scenes, but the data completeness is poor due to blocking and missing information
Solution Approach 1:
The patent applies partial action by processing only the most relevant point cloud clusters that correspond to actual obstacles, rather than processing all point cloud data. By filtering and selecting significant clusters based on spatial distribution and motion characteristics, the system maintains 3D scene handling capability while reducing the impact of incomplete or blocked data
Solution Approach 2:
The patent implements feedback mechanisms where the system continuously monitors the quality and completeness of point cloud data, and adjusts its processing strategy accordingly. When data completeness deteriorates due to blocking, the system can switch to alternative tracking methods or increase reliance on predictive models, thereby maintaining adaptability to complex scenes despite information loss
3Productivity
If traditional centroid-based velocity calculation is used, then the computation is fast, but the reliability is low due to large errors from blocking and position fluctuations
Solution Approach 1:
The patent performs preliminary actions by pre-processing point cloud data to identify and filter out unreliable clusters before velocity calculation. By preparing and organizing point cloud data into meaningful clusters in advance, the system ensures that only high-quality data is used for velocity computation, thereby improving reliability without significantly reducing computation speed
Solution Approach 2:
The patent substitutes the simple mechanical centroid-based calculation with a more sophisticated cluster-based approach that incorporates multiple factors such as spatial distribution, motion consistency, and data quality metrics. This substitution increases computational complexity slightly but dramatically improves reliability by replacing direct centroid differentiation with a multi-criteria evaluation system
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
Embodiments of the present disclosure provide a method and apparatus for determining a velocity of an obstacle, a device, a medium and a computer program product. A specific implementation solution includes: acquiring a first point cloud data of the obstacle at a first time and a second point cloud data of the obstacle at a second time; registering the first point cloud data and the second point cloud data by moving the first point cloud data or the second point cloud data; and determining a moving velocity of the obstacle based on a distance between two data points in a registered data point pair.