Vehicle Type Recognition Using PCA and EM Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cluster analysis methods, such as X-means clustering, face challenges in accurately identifying the number of clusters and suffer from local mode issues when dealing with vehicle data for traffic monitoring and maintenance cost forecasting, often misassigning data points due to the use of Euclidean distance.
Innovation Solution
The implementation of the Expectation-Maximization (EM) algorithm with Mahalanobis distance to improve cluster analysis, allowing for the merging of similar clusters based on Bayesian Information Criterion (BIC) scores, and using Principal Component Analysis (PCA) to split and reassign data points, thereby enhancing the accuracy of vehicle type identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If X-means clustering is used to detect the number of clusters automatically, then the need for user input is reduced, but local mode problems become more severe
Solution Approach 1:
The patent combines multiple clustering algorithms (K-means, X-means) with model selection criteria (BIC, AIC) to create a hybrid approach. This merging allows the system to automatically detect the number of clusters while using statistical criteria to evaluate and select the optimal clustering solution, thereby reducing local mode problems and improving reliability.
Solution Approach 2:
The patent implements feedback mechanisms by using information criteria (BIC, AIC) to evaluate clustering results and guide further clustering iterations. The system uses the evaluated results to adjust clustering parameters and re-run clustering, creating a feedback loop that improves accuracy while maintaining automatic operation.
2Ease of manufacture
If Euclidean distance is used for clustering, then the calculation is simple, but data points are misassigned to wrong centroids
Solution Approach 1:
The patent changes the distance metric parameter from Euclidean distance to Mahalanobis distance. This parameter change accounts for correlations between variables and different scales, improving measurement precision in data point assignment while maintaining computational feasibility through efficient algorithms.
3Ease of manufacture
If K-means clustering is used, then the implementation is simple, but the user must supply the number of clusters
Solution Approach 1:
The patent creates a universal clustering system that incorporates multiple algorithms (K-means for simple cases, X-means for automatic detection) and multiple model selection criteria (BIC, AIC). This multi-functional system can automatically determine the number of clusters using information criteria while still allowing simple K-means to be used when the number of clusters is known, thereby satisfying both simplicity and automatic operation requirements.
Data Source
AI summary
A method is provided herein for determining and recognizing types of vehicles passing a check point. The method takes advantage of an EM algorithm which is up-loaded into a CPU and which processes data of the vehicles which drive past a checkpoint, the data being representative of essential characteristics of vehicles to produce an output model of the traffic volumes of the various types of vehicles. This model enables the forecasting of future road maintenance costs and the planning and designing of future road networks.