Adaptive In-Vehicle Lidar Downsampling for Sparse-Object Areas
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing down-sampling methods for vehicle position estimation using lidar point cloud data reduce accuracy and robustness in areas with few objects due to excessive reduction in the number of data points.
Innovation Solution
An information processing device that adaptively changes the size of down-sampling based on the number of associated measurement points, ensuring the number remains within a target range to optimize data usage for accurate position estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If down-sampling process is applied to reduce the total number of point cloud data, then calculation time is suppressed and self position estimation accuracy is improved, but the number of point cloud data to be used for calculation is greatly reduced in areas with few objects, lowering accuracy and robustness
Solution Approach 1:
The patent applies dynamic down-sampling by adjusting the down-sampling intensity based on the density of measurement points in different spatial regions. In areas with high point cloud density, stronger down-sampling is applied to reduce data quantity and calculation time. In areas with low point cloud density, weaker down-sampling is applied to maintain sufficient data points for accurate position estimation, thus resolving the contradiction between estimation accuracy and data quantity dynamically
Solution Approach 2:
The patent implements local quality by applying different down-sampling strategies to different spatial regions based on their characteristics. The system divides the measurement space into regions with different point cloud densities and applies appropriate down-sampling factors to each region, ensuring that data quantity is optimized locally rather than uniformly across the entire space
2Loss of time
If down-sampling process is applied to equalize dense point cloud data, then calculation time is suppressed, but the robustness of self position estimation is lowered in areas with few objects
Solution Approach 1:
The system dynamically adjusts the down-sampling factor based on real-time analysis of point cloud density distribution. By monitoring the density of measurement points in different regions, the system adapts the down-sampling intensity to maintain robustness in sparse areas while achieving time savings in dense areas
Solution Approach 2:
The patent changes the down-sampling parameter (down-sampling factor or threshold) based on the detected point cloud density characteristics. By adjusting this parameter dynamically, the system optimizes the balance between calculation speed and estimation robustness according to the specific spatial conditions
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The controller 13 of the in-vehicle device 1 acquires point cloud data outputted by a lidar 2. Then, the controller 13 generates processed point cloud data obtained by down-sampling the point cloud data. The controller 13 matches the processed point cloud data with voxel data VD which represents a position of an object with respect to each voxel that is a unit area and thereby associates a measurement point of the processed point cloud data with the each voxel. Then, the controller 13 changes the size of the subsequent down-sampling based on the associated measurement points number Nc which is the number of measurement points associated with the each voxel among the measurement points of the processed point cloud data.