High Precision Map Generation Using 3D Laser Point Cloud Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing electronic maps have low precision and lack essential dimensionality, such as road lane marking, height, slope, curvature, and border information, requiring extensive manual updates and inefficient data generation.
Innovation Solution
A method and apparatus utilizing 3D laser point cloud data and machine learning algorithms to generate high precision maps by acquiring and processing 3D laser point cloud data, determining position information, rendering pixel points, identifying and clustering traffic information, and loading it into a grid map to include road shape, slope, curvature, and border details.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If satellite surveying and mapping is used to obtain raw map data, then map coverage is achieved, but map precision is low with errors in meters or tens of meters
Solution Approach 1:
The patent transitions from traditional 2D satellite imagery to 3D point cloud data representation, adding vertical dimension information for height, slope, and curvature data. This dimensional enhancement enables precise capture of road surface characteristics and three-dimensional terrain features that were previously unavailable in conventional maps.
Solution Approach 2:
The patent replaces manual mechanical surveying and periodic field updates with automated laser radar scanning and machine learning algorithms. The system automatically extracts road features, lane markings, and traffic information from point cloud data, eliminating the need for manual data collection and processing while achieving centimeter-level precision.
2Productivity
If manual operations are used to update navigation map information, then map accuracy can be maintained, but data generating efficiency is low
Solution Approach 1:
The system implements self-service through automated machine learning algorithms that independently process laser radar point cloud data, extract road features, identify lane markings, and generate updated map information without human intervention. The algorithms automatically cluster point cloud data, recognize traffic signs, and produce high-precision map outputs, enabling continuous autonomous map updates.
Solution Approach 2:
The patent changes the fundamental parameters of map data collection by using laser radar to capture millions of three-dimensional points with precise spatial coordinates and intensity values. This parameter transformation from 2D pixel data to 3D point cloud data enables automated feature extraction and significantly improves both processing efficiency and measurement precision.
3Loss of information
If traditional electronic map data is used, then basic navigation functionality is provided, but essential dimensionality such as road lane marking, height, slope, curvature, and border information is absent
Solution Approach 1:
The patent segments the complex point cloud data into distinct feature categories including road surfaces, lane markings, curbs, barriers, and traffic signs. The machine learning system separately processes and identifies each feature type through specialized algorithms, then integrates them into a comprehensive map structure that preserves all dimensional information without overwhelming processing complexity.
Data Source
AI summary
The present application discloses a method and an apparatus for generating a high precision map. According to an embodiment, the method comprises: acquiring three-dimensional (3D) laser point cloud data and information related to a grid map for generating the high precision map; determining position information of each piece of the 3D laser point data of the 3D laser point cloud data in the grid map; rendering each pixel point in the grid map, by using the reflection value of corresponding 3D laser point data of the 3D laser point cloud data, in order to generate each of grid images in the grid map; identifying, by using a machine learning algorithm, traffic information of each of the grid images in the grid map; clustering the traffic information of each of the grid images in the grid map to obtain the traffic information of the grid map; and loading the traffic information of the grid map into the grid map to generate the high precision map. The embodiment implements the generating of a high precision map with a high precision and a plurality of dimensions.