Hierarchical Classification Based on Conditional SOM for Situation Analysis

Abstract

This paper provides an approach of hierarchical classification based on Conditional Self-Organizing Map (CSOM) algorithm in order to reduce the risk of collision on the road. Two types of data (using Dot Distribution Representation) have been taken for hierarchical classification. First type is Cross shape road data and second one is object data (in this case coordinates of vehicles as object data are considered) on the road. The purpose of hierarchical classification of road-vehicle situation (i.e. first classification of cross shape road and second classification is group of number of vehicles corresponding to cross shape road) is to anticipate the density of available objects on the road. This estimated information can be used by driving systems and helpful to take decision such as by reducing speed of vehicle or by changing lane or by changing the direction of vehicle in order to reduce the risk of collision.