상대 차량의 움직임 판단을 위한 특징벡터 커널링 기반 멀티클래스 서포트 벡터 머신

Abstract

In this paper, we propose a feature vector kernelling method for the multi-class support vector machine (MSVM) in the object vehicle motion classification. Since the conventional radar system has the limitation of low resolution of lateral information, it is difficult to be utilized in advanced driver assistant systems. The MSVM is used to resolve this problem in an object vehicle motion classification problem. For the feature vector of the MSVM, we present a feature vector kernelling method using Kalman filter. The proposed algorithm was validated with a data set not included in the training data set. From the experimental results, we observed that the proposed system can identify the cut-in/out situation.