LIDAR Depth Map Feature Extraction for Lane-Level Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional GPS, WiFi, and Bluetooth-based vehicle localization methods are imprecise due to multi-pathing, occlusion, and lack of precision in three-dimensional space, failing to provide lane or road-level positioning.
Innovation Solution
A fingerprint database is developed using two-dimensional feature geometries extracted from depth maps by LIDAR data, allowing end-user devices to determine their location by matching surrounding features with encoded fingerprints, eliminating the need for expensive graphics processing units and enabling precise positioning even without conventional geo-positioning technologies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional GPS, WiFi, and Bluetooth methods are used for vehicle localization, then the system can provide basic positioning functionality, but the positioning precision deteriorates and cannot achieve lane or road-level accuracy
Solution Approach 1:
The patent replaces conventional radio wave-based positioning systems (GPS, WiFi, Bluetooth) with a visual positioning system using LIDAR depth maps and image processing. This substitution transitions from electromagnetic wave measurement to optical measurement, achieving lane-level precision by extracting geometric features from visual data rather than relying on unreliable signal strength or timing information
Solution Approach 2:
The patent creates a digital fingerprint database that copies and stores the geometric features of road segments from LIDAR depth maps. By comparing real-time extracted features against these stored fingerprints, the system achieves precise localization without needing direct signal measurements from transmitting stations, thereby overcoming signal reliability issues
2Measurement precision
If LIDAR depth maps and feature extraction are used for positioning, then positioning precision improves to lane or road-level accuracy, but device complexity increases
Solution Approach 1:
The patent extracts only the essential geometric features (edges, corners, lines) from LIDAR depth maps to create positioning fingerprints. By taking out only the necessary geometric information rather than processing complete 3D point clouds, the system achieves lane-level precision while reducing computational complexity and enabling implementation on cost-effective hardware
Solution Approach 2:
The patent segments the LIDAR depth map data into discrete geometric features (edges, corners, lines) that can be independently extracted and matched. This segmentation transforms complex continuous 3D data into discrete 2D geometric primitives, simplifying the matching process against the fingerprint database while maintaining positioning accuracy
3Device complexity
If conventional GPS and signal-based methods are used, then the system architecture remains simple, but positioning precision deteriorates and cannot achieve lane-level accuracy
Solution Approach 1:
The patent transitions from 3D spatial signal propagation measurement to 2D geometric feature extraction and matching. By projecting 3D LIDAR depth map data onto 2D images and extracting geometric features in the 2D domain, the system achieves higher positioning precision while simplifying the computational architecture compared to 3D point cloud processing
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This method provides improved positioning accuracy and reduces computation costs, enabling real-time location determination with low-cost visualization technology, and can function in the absence of conventional GPS, achieving lane or road-level precision.
Implementation Method 1
A processor of the end-user device extracts two-dimensional feature geometries from a depth map collected by a light detection and ranging (LIDAR) device
Implementation Method 2
extracts two-dimensional feature geometries from a depth map collected by a light detection and ranging (LIDAR) device
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method of localisation of a device using multilateration comprising collecting, by an end-user device, a depth map at a location in a path network (s101); extracting, using a processor of the end-user device, two-dimensional feature geometries from the depth map (s103); identifying a number of control points in the extracted feature geometries (s105); calculating distances between the end-user device and the identified control points (s107); receiving location reference information for each identified control point from an external database (s109); and determining a geographic location of the end-user device in the path network through a multilateration calculation using the location reference information of the identified control points and the calculated distances between the end-user device and each identified control point (s111).