Road Object Positioning via Density-Based Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing GPS-type geolocation systems for road objects are imprecise, leading to uncertainty in the position of road objects, and some detections result in false positives, which is unacceptable for applications like driving assistance and autonomous vehicle navigation.
Innovation Solution
A system and method utilizing unsupervised classification based on distribution density to process geolocated traffic information from multiple road vehicles, creating characteristic vectors and calculating central tendency measures to accurately determine the position of road objects, incorporating DBSCAN, OPTICS, CLARANS, DENCLUE, and CLIQUE algorithms, and separating coordinates into latitude, longitude, and azimuth components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If GPS-type geolocation systems are used to locate road objects, then the system is simple and easy to implement, but the positioning accuracy is imprecise and produces false positives
Solution Approach 1:
The patent combines data from multiple sources including GPS coordinates, road vehicle sensor data, and unsupervised machine learning algorithms to create a more accurate positioning system. By merging these diverse data sources and processing them through density-based clustering, the system achieves higher measurement precision while distributing the complexity across multiple integrated components rather than relying on a single simple GPS system
2Reliability
If data from multiple road vehicles is collected and processed using unsupervised machine learning, then the positioning accuracy improves, but the processing complexity and computational requirements increase
Solution Approach 1:
The system employs unsupervised machine learning algorithms that automatically process and classify data without requiring manual intervention or pre-labeled training data. The density-based clustering algorithms self-organize the geolocated data points from multiple vehicles, automatically identifying patterns and reducing false positives. This self-service approach improves detection reliability while keeping the processing complexity manageable through automated rather than manual methods
3Measurement precision
If unsupervised classification algorithms are applied to partition characteristic vectors, then false positives are reduced and accuracy improves, but the computational time and processing steps increase
Solution Approach 1:
The patent segments the processing into distinct phases: first partitioning geolocated data points into characteristic vectors, then applying unsupervised classification to partition these vectors into classes, and finally calculating central tendency measures for each class. This segmentation of the processing workflow allows the system to apply complex algorithms in a structured manner, improving coordinate precision through multiple processing stages rather than a single time-consuming operation
Data Source
Figure 1~2
Figure 3
AI summary
The present invention concerns systems, methods and a set of computer programs for accurately locating the position of at least one road object associated with a portion of a road network. The general principle of the invention is based on determining the position of a road object using unsupervised classification based on distribution density. In the invention, the classification is applied in two phases. In the first phase, it is applied a first time to the plurality of geographical coordinates associated with a road object so as to automatically group it into homogeneous classes. In the second phase, it is applied a second time to each class produced in the first phase, so as to automatically group it into homogeneous sub-classes based on the azimuth angles of the road object that are associated with the geographical coordinates of the class.