A method combines Top-hat Transformation, Morphological Gradient and Background Subtraction is presented in this paper to solve the question of cast shadows and split of vehicle. The method adopts top-hat transformation both on input image and background image to remove the shadows and detect road lines respectively. Then morphological gradient is obtained by multiple structuring elements. The final modified background subtraction which performs on the binary images removes the road lines which impact the result much. Different from other methods, the proposed method eliminates the shadows and extracts the whole vehicle even if the color of the vehicle is similar to the road and the contour beside windshield is very weak. Experimental results prove that the method works well under different vehicle colors and sizes, comparing with the simple background subtraction and the background subtraction based on top-hat transformation. The error rate indicates the improvement.