LIDAR Point Cloud Merging with Normal Covariance Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for processing LIDAR sensor data, such as the Iterative Closest Point (ICP) algorithm, face challenges in achieving accurate convergence and object representation due to limited error metric minimization, leading to distorted objects in comprehensive scene representations.
Innovation Solution
A modified ICP framework is implemented, including initialization, filtration, matching, transformation estimation, and termination criteria steps, which employs a Normal Covariance Filter to account for measurement errors and geometric criteria to improve data point selection and filtering, enhancing the accuracy of data merging and object detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the standard ICP algorithm is used to merge 3D point clouds, then the merging process can be completed, but the convergence accuracy is limited and objects may appear distorted in the comprehensive scene representation
Solution Approach 1:
The patent modifies the ICP algorithm by changing the error metric from traditional point-to-point or point-to-plane distances to a normal covariance-based metric. This parameter change in the optimization objective function allows the algorithm to account for measurement uncertainties and geometric constraints, improving convergence accuracy and preventing object distortion in the merged point cloud representation
Solution Approach 2:
The patent replaces the traditional geometric distance minimization mechanism with a statistical approach using normal covariance matrices. Instead of relying solely on Euclidean distances between points, the new mechanism incorporates probabilistic information about point uncertainties and surface orientations, leading to more reliable convergence and accurate object representation
2Quantity of substance
If more LIDAR data points are included in the merging process, then the comprehensive scene representation becomes more detailed, but the computational burden increases
Solution Approach 1:
The patent extracts and utilizes normal covariance information from LIDAR measurements to create a more efficient error metric. By incorporating this extracted statistical information into the optimization process, the algorithm achieves better convergence with fewer iterations, reducing computational burden while maintaining detailed scene representation from all LIDAR data points
Solution Approach 2:
The patent performs preliminary calculation of normal covariance matrices for each LIDAR point cloud before the merging process. This preliminary action prepares the data in a form that enables more efficient optimization during the ICP algorithm, reducing the computational burden of the merging process while preserving all detailed information from the original data points
Data Source
AI summary
Method and device for processing LIDAR sensor data are disclosed. The method includes: receiving a first dataset and a second dataset having pluralities of data points; matching at least some of the plurality of first points with at least some of the plurality of second points, thereby determining a plurality of pairs; for the given one of the plurality of pairs, determining a pair-specific filtering parameter by calculating neighbour beam distances between the given first data point and respective ones the set of neighboring points, a given neighbour beam distance being representative of a linear distance between the given first data point and a respective one of the set of neighbouring points; in response to the pair-specific parameter being positive, excluding the given one of the plurality of pairs from further processing; and processing the reduced plurality of pairs for merging the first dataset and the second dataset.