No abstract

No abstract

Abstract. Due to the influence of gray contrast ratio on background image with the target vehicle, traditional vehicle detection algorithms cannot adapt to complicated traffic environment. This paper proposes a new vehicle classification method based on locational matrix and minimum of local edge region. First, the lane line is determined by gray gradient variation and the regions surrounded by lane lines is regarded as the recognition region for further vehicle detection. Second, grid partition of the original image is conducted and classification label of each grid is determined by the assumed gray threshold. Then locational matrix is obtained based on the label number of the corresponding grids and the similarity rate is analyzed with different ε to get the optimal ε value. The relationship between the distance from the mean of local region gray to the connected domain of vehicle body and ε is the iteration termination condition. Finally, vehicle outline edge is refined through the minimum of edge local region. Test results show that our method classifies vehicle outline edge accurately, compared with single image match method.