3D Point Cloud Ground Classification via Cell-Based Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for processing 3D point clouds from vehicle sensors, such as LIDAR or radar, face challenges in accurately differentiating ground points from object points, especially in varying environments like hilly or flat terrains, leading to potential misclassification and increased computational requirements.
Innovation Solution
A method that divides the reference plane into cells, identifies starting ground points, and uses regression analysis to approximate cell planes, iteratively processing candidate cells to accurately map ground profiles, while also classifying points as ground or object points based on predefined criteria, thereby reducing computational demands and improving reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If current methods process 3D point clouds to differentiate ground points from object points, then ground surface identification is achieved, but misclassification occurs and computational requirements increase
Solution Approach 1:
The reference plane is divided into multiple cells, and the point cloud is processed cell by cell. Each cell is evaluated independently to determine if it contains ground points, object points, or mixed points. This segmentation approach reduces computational complexity by breaking down the large-scale classification problem into smaller, manageable sub-problems, while improving reliability through localized analysis that adapts to terrain variations in different regions
Solution Approach 2:
The method applies different processing strategies to different cells based on their local characteristics. Cells are classified as ground cells, object cells, or mixed cells, and each type receives appropriate handling. This local quality approach allows the system to adapt to varying terrain conditions (flat vs. hilly) in different regions, improving overall classification accuracy without uniformly increasing computational requirements across the entire point cloud
2Quantity of substance
If the first predefined distance is extended to cover larger areas, then more ground points are captured, but ground surface curvature deviations increase
Solution Approach 1:
By dividing the reference plane into multiple cells, the system can extend the first predefined distance to cover larger areas and capture more ground points, while maintaining measurement precision through localized cell-based processing. Each cell independently evaluates ground point criteria, ensuring that even in regions with ground surface curvature, the local terrain characteristics are accurately captured without being affected by deviations in distant regions
Solution Approach 2:
The method transitions from a global distance-based filtering approach to a cell-based dimensional framework. Instead of applying a single distance threshold to the entire point cloud, the system evaluates points within the context of their specific cell location, adding the spatial dimension of cell membership to the classification process. This allows extended coverage area while maintaining precision through localized evaluation
Data Source
AI summary
A method and to a device for processing a 3D point cloud representing surroundings, which is generated by a sensor. Initially, starting cells are identified based on ascertained starting ground points within the 3D point cloud which meet at least one predefined ground point criterion with respect to a reference plane divided into cells. Thereafter, cell planes are ascertained for the respective starting cells of the reference plane. Thereafter, estimated cell planes and ground points are ascertained for candidate cells deviating from the starting cells based on the cell planes of the starting cells, which are subsequently converted into final cell planes. As a result of such a cell growth originating from the starting cells, the cells of the reference plane are iteratively run through and processed so that the 3D point cloud is reliably classifiable into ground points and object points based on this method.