Geometric Fingerprinting for Lane-Level Vehicle Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle localization methods using GPS, WiFi, and Bluetooth are imprecise due to multi-pathing, occlusion, and lack of precision in location references, making it difficult to achieve lane-level or road-level positioning.
Innovation Solution
Development of a fingerprint database using depth maps from depth sensing cameras to identify and encode two-dimensional feature geometries from physical structures in a path network, allowing devices to determine their geographic location by comparing extracted features with the database.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If GPS, WiFi, or Bluetooth are used for vehicle localization, then positioning can be achieved, but precision is insufficient to reach lane-level or road-level accuracy
Solution Approach 1:
The patent replaces GPS satellite-based positioning, WiFi signal-based localization, and Bluetooth signal-based positioning with a vision-based geometric fingerprinting system. Instead of relying on radio wave propagation and signal triangulation, the system uses depth sensing cameras to capture geometric features of the physical environment and matches them against pre-stored fingerprint databases, achieving lane-level positioning accuracy through geometric pattern recognition rather than signal-based methods
Solution Approach 2:
The patent creates geometric fingerprints by capturing and storing three-dimensional depth maps of the environment that represent the unique geometric configuration of physical structures at specific locations. These depth map copies are then matched against live captures to determine position, replacing the need for signal-based location references with geometric pattern matching
2Measurement precision
If conventional positioning methods are used, then location can be determined, but computational costs are high requiring expensive graphics processing units
Solution Approach 1:
The patent extracts only the essential geometric features from depth maps - specifically two-dimensional feature geometries representing physical structures like buildings, trees, and other environmental elements. By extracting only these critical geometric patterns rather than processing entire high-resolution images or performing complex signal processing calculations, the system achieves accurate positioning with significantly reduced computational requirements that eliminate the need for expensive GPUs
3Ease of operation
If signal-based localization (WiFi, Bluetooth) is used, then positioning is possible, but precision is degraded due to occlusion and lack of precision in transmitting station locations
Solution Approach 1:
The patent substitutes signal-based localization methods with vision-based geometric feature matching. Instead of relying on WiFi or Bluetooth signals that are susceptible to occlusion and imprecise transmitter location knowledge, the system uses depth sensing cameras to capture and match geometric patterns of the environment, which are not affected by signal occlusion and provide precise location information through geometric correspondence
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Systems, apparatuses, and methods are provided for developing a fingerprint database for and determining the geographic location of an end-user device (e.g., vehicle, mobile phone, smart watch, etc.) with the database. A fingerprint database may be developed by receiving a depth map for a location in a path network, and then identifying physical structures within the depth map. The depth map may be divided, at each physical structure, into one or more horizontal planes at one or more elevations from a road level. Two-dimensional feature geometries may be extracted from the horizontal planes. At least a portion of the extracted feature geometries may be encoded into the fingerprint database.