Vehicle Class Identification Under Cluttered Background Using Statistical Features and Correlation Technique

Vehicle class identification is an important task in pattern recognition. Identifying vehicle with cluttered background affects accuracy of the overall system. Removal of cluttered background from images gives better results in vehicle class identification. In this paper, the cluttered background and mild occlusions are removed from the static images. Statistical features are extracted from the background removed images. The statistical feature of master image is correlated with the features of images from the standard UIUC (University of Illinois, Urbana-Champaign) database. This paper addresses the issues to identify vehicle class of real-world images containing side views of cars class with that of non-cars class. Critical evaluation of the proposed method has improved to an accuracy of 91.7%.