Resolution-Adaptive Point Cloud Fusion for Vehicle Obstacle Sensing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing obstacle detection systems for autonomous vehicles do not effectively account for resolution differences between various 3D scanning sensors, leading to poor performance in obstacle detection and navigation due to uncompensated low resolution in certain dimensions.
Innovation Solution
A resolution-adaptive fusion method that combines point clouds from different 3D scanning sensors by generating volumetric surface functions and forming a composite surface function, which compensates for poor resolution by incorporating data from sensors with better resolution, enhancing detection capabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If point clouds from multiple sensors with different resolutions are fused using existing methods, then the quantity of data is increased, but the measurement precision deteriorates due to unresolved resolution differences
Solution Approach 1:
The patent transforms point cloud data from different resolution sensors into a common coordinate system with unified resolution by applying resolution transformation. This involves resampling points from high-resolution sensors to match the resolution of lower-resolution sensors, ensuring that all point clouds contribute equally to the fused result without introducing resolution mismatches that would degrade measurement precision.
Solution Approach 2:
The patent creates a composite point cloud by fusing data from multiple sensor types (e.g., LIDAR, radar, vision sensors) with different resolution characteristics. By transforming and harmonizing the resolution parameters of each sensor's point cloud before fusion, the system achieves a composite data structure that maintains high measurement precision while leveraging the complementary strengths of multiple sensors.
2Loss of information
If sensor fusion is performed at the point cloud level by registering and aligning point clouds, then the completeness of environmental representation is improved, but the manufacturing precision deteriorates due to lack of resolution compensation
Solution Approach 1:
The patent applies resolution transformation as a parameter change operation that modifies the spatial distribution and density of points in each point cloud before fusion. This ensures that when point clouds are registered and aligned in the common coordinate system, they all share consistent resolution characteristics, preventing degradation of precision while maintaining complete environmental representation.
Solution Approach 2:
The patent performs resolution transformation and harmonization as a preliminary action before the actual point cloud fusion process. By pre-processing the point clouds to ensure uniform resolution, the system avoids precision loss during the subsequent registration and alignment operations, while still achieving complete environmental coverage through multi-sensor fusion.
3Device complexity
If existing sensor fusion methods are used that process raw point clouds separately and independently, then the device complexity is reduced, but the measurement precision deteriorates due to failure to account for resolution differences
Solution Approach 1:
The patent introduces resolution transformation as an additional processing step that modifies the resolution parameters of point clouds from different sensors. While this increases processing complexity compared to simple independent processing, it enables accurate obstacle detection by ensuring that resolution differences between sensors do not degrade the precision of the fused results.
Data Source
AI summary
A method for obstacle detection and navigation of a vehicle using resolution-adaptive fusion includes performing, by a processor, a resolution-adaptive fusion of at least a first three-dimensional (3D) point cloud and a second 3D point cloud to generate a fused, denoised, and resolution-optimized 3D point cloud that represents an environment associated with the vehicle. The first 3D point cloud is generated by a first-type 3D scanning sensor, and the second 3D point cloud is generated by a second-type 3D scanning sensor. The second-type 3D scanning sensor includes a different resolution in each of a plurality of different measurement dimensions relative to the first-type 3D scanning sensor. The method also includes detecting obstacles and navigating the vehicle using the fused, denoised, and resolution-optimized 3D point cloud.