Centroid-Based Mean Shift for Crowdsourced Road Object Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing clustering algorithms for crowdsourced map data, such as K-means, spectral clustering, and mean shift, struggle with selecting parameters, sensitivity to noise, uneven data distribution, and poor performance in complex road environments, leading to inefficient and inaccurate clustering of road object elements.
Innovation Solution
A clustering method that transforms polygonal road object data into centroids, using geometric relationships like centroid distance, intersection area ratio, and minimum to maximum area ratio, and incorporates a mean shift algorithm to handle abnormal points and diverse data distributions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If K-means algorithm is used for clustering road object data, then the clustering process is simple, but it is difficult to select K-value and converges poorly for non-convex data sets with imbalanced data volumes
Solution Approach 1:
The patent transforms the fixed K-value requirement of K-means into a dynamic parameter selection process. It uses density-based parameters (eps, minPts) instead of predetermined cluster numbers, allowing the algorithm to adapt to varying data distributions and automatically determine the appropriate number of clusters based on data characteristics rather than requiring manual K-value selection.
Solution Approach 2:
The patent introduces dynamic adaptability to clustering by making the algorithm sensitive to local density variations. The cluster formation process dynamically adjusts to data distribution patterns, allowing clusters to form organically based on density thresholds rather than forcing data into predetermined convex shapes, thus handling non-convex and imbalanced data effectively.
2Adaptability or versatility
If spectral clustering algorithm is used, then clustering can be performed on complex data distributions, but the construction of similarity matrices is time-consuming and computing power is weak for large data volumes
Solution Approach 1:
The patent extracts only the essential density information from the data without constructing complete similarity matrices. By focusing on local density estimation through distance calculations to nearest neighbors rather than computing all pairwise similarities, it captures the necessary structural information for clustering while dramatically reducing computational complexity from O(n²) to O(n log n) or O(n).
Solution Approach 2:
The patent uses approximate distance calculations and density estimations that are computationally inexpensive rather than expensive exact similarity measurements. The algorithm accepts approximate density values and distance metrics that can be computed efficiently, trading minor precision loss for significant gains in processing speed and scalability to large datasets.
3Reliability
If DBSCAN clustering algorithm is used, then density-based clustering can be achieved, but clustering convergence time is long for large sample sets
Solution Approach 1:
The patent segments the clustering process into distinct phases: density estimation, cluster seed identification, and point assignment. By dividing the computation into these stages and using spatial indexing or sorted data structures during density estimation, it reduces the overall time complexity while maintaining density-based clustering quality, avoiding the O(n²) bottleneck of traditional DBSCAN.
Solution Approach 2:
The patent performs preliminary sorting or indexing of data points by distance or density metrics before the main clustering loop. This preprocessing step allows for efficient nearest neighbor queries and density calculations during cluster formation, avoiding repeated expensive distance computations and significantly reducing convergence time for large datasets.
4Ease of operation
If conventional mean shift clustering algorithm is used, then clustering can be performed, but it uses only distance for clustering determination and cannot comprehensively consider actual complex road environments
Solution Approach 1:
The patent combines multiple clustering criteria into a composite density metric. Instead of relying solely on Euclidean distance, it integrates distance with density information and potentially other domain-specific features to form a comprehensive similarity measure. This composite approach enables the algorithm to capture complex spatial relationships in road environment data while maintaining the iterative mean shift framework.
Solution Approach 2:
The patent replaces the simple distance-based attraction mechanism of conventional mean shift with a density-aware attraction mechanism. Instead of moving points purely toward the mean based on Euclidean distance, points are attracted toward density centers weighted by local density estimates, allowing the algorithm to adapt to varying data densities and complex spatial patterns in road environments.
Data Source
Figure 1
Figure 2
AI summary
Disclosed are a clustering method and system for road object elements of a crowdsourced map, and a storage medium. Data of road objects are skillfully represented by centroids thereof according to an improved mean shift algorithm, and geometric relationships between different objects are selected according to different types of road objects as measurement conditions for clustering. Neighborhood relationships of the objects are considered in the clustering process according to the characteristics of road element data types, and clustering output of different road element types is ultimately obtained. The method skillfully transforms object data into their centroids as clustering input, and flexibly uses multiple geometric relationships of the objects according to different scenarios, so the method has the advantages of conventional mean shift algorithms, such as high efficiency, small computation quantity, and insusceptibility to abnormal points. The method further incorporates the neighborhood relationships of the objects to solve the shortcoming that the conventional mean shift algorithms are not suitable for complex data distribution types.