Cluster and Discriminant Analysis for Vehicles Detection

It is very useful to build an automatic computer system to recognize the types of vehicles passing a checkpoint given some easy-to-get data about the vehicles, such as the distances between axles, the weights on each axle. Such a system has many applications, for example, in monitoring traffic volumes and identifies the type of vehicle, which will be helpful in budgeting road maintenance costs. The main goal of this project is to develop a better methodology for cluster analysis with application to the vehicle detection problem. The simplest clustering technique is the K-means clustering. However, K-means clustering requires that the users supply with a number of clusters. X-means clustering may be an alternative method since it can detect the number of clusters with some simple criteria. But X-means would introduce more severe local mode problem. We will investigate a new method to overcome the problem by merging similar clusters after running X-means clustering.

Faculty Supervisor:

Drs. William Laverty and Longhai Li

Student:

Zhengrong Li

Partner:

International Road Dynamics Inc.

Discipline:

Mathematics

Sector:

Automotive and transportation

University:

University of Saskatchewan