LiDAR Ground Classification Using Terrain-Adaptive Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing ground classification algorithms for LiDAR point clouds are terrain-specific, require manual parameter tuning, and lack consideration of intensity information, leading to inaccuracies and inefficiencies in classifying ground points.
Innovation Solution
A machine learning model trained on features derived from conventional ground classification algorithms, point cloud attributes, and contextual features, providing ground classification probabilities without manual parameter tuning, improving accuracy and scalability across various terrains.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional ground classification algorithms are used, then ground points can be classified, but the algorithms are terrain-specific and require manual parameter tuning, reducing adaptability and ease of operation
Solution Approach 1:
The machine learning model performs self-adjustment by automatically learning optimal classification parameters from training data across multiple terrains, eliminating the need for manual parameter tuning. The model adapts to different terrain types through automated training processes rather than requiring human intervention for each terrain type.
Solution Approach 2:
A single machine learning model is designed to handle multiple terrain types universally, replacing the need for separate terrain-specific algorithms. The model is trained on diverse terrain data and can generalize its classification capabilities across varied environments without requiring terrain-specific configuration.
2Measurement precision
If conventional ground classification algorithms are used, then ground points can be classified, but intensity information is not considered, reducing measurement precision
Solution Approach 1:
The patent combines multiple data sources including intensity information, spatial coordinates, and classification results from multiple algorithms into a unified feature set for the machine learning model. This merging of previously underutilized intensity data with other features improves overall classification precision.
3Measurement precision
If multiple ground classification algorithms are applied, then classification accuracy can be improved, but computational time and complexity increase
Solution Approach 1:
Multiple ground classification algorithms are applied in advance during the training phase to generate labeled training data and initial classifications. The machine learning model learns from these pre-computed results, so during actual operation, only the trained model needs to run, significantly reducing computational time compared to running multiple algorithms for each classification task.
4Measurement precision
If machine learning model is trained on comprehensive features, then classification accuracy is improved, but device complexity and training requirements increase
Solution Approach 1:
The feature extraction process is segmented into distinct components: point cloud features, cell features, and regional features are extracted and processed separately before being combined for model training. This segmentation allows for systematic development and debugging of each feature type independently, reducing overall training complexity.
Data Source
AI summary
Example systems, methods, and non-transitory computer readable media are directed to obtaining a point cloud that represents an environment based at least in part on a plurality of points in three-dimensional space; determining corresponding classifications of points in the point cloud as ground or not-ground based at least in part on a plurality of ground classification algorithms; determining respective point cloud features associated with the points in the point cloud; determining respective cell features associated with a plurality of cells that segment the point cloud; generating feature data for a machine learning model based at least in part on one or more of: the classifications of the points based on the plurality of ground classification algorithms, the point cloud features, or the cell features; and classifying the points in the point cloud based at least in part on an output from the machine learning model.