Subspace-based preceding vehicle detection

Abstract

In this paper, a vision-based preceding vehicle detection scheme using the statistical information of the vehicles and non-vehicles obtained is presented. K clusters are created using the simple K -means clustering algorithm. The partitions are recomputed using the nested subspacing concept. Mahalanobis distance based measure is used for grouping the image patterns and recognizing the vehicles. The performance of the proposed vehicle detection scheme is compared with that of Multi-Clustered Modified Quadratic Discriminant Function (MC-MQDF) method of preceding vehicle detection. Experimental results prove that the proposed scheme is more suitable for a reliable driver assistance system.