Vehicle face detection benefits a wide range of applications such as vehicle identification and semantic segmentation of the vehicle. Although great  dedication to this task, most existing researches focus on the detection and positioning of the entire area of the vehicle face. We propose a new vehicle face detection method based on selective search. The approach has two steps. Firstly, the vehicle image is denoised through Gaussian filter as the preprocess. Secondly, we use the image segmentation algorithm based on graph presentation to obtain the initial image segmentation regions from the preprocessed images. We calculate the similarity degree of adjacent regions in color, texture, size and coincidence degree, and then merge the initial split area with the one with the highest similarity. We conduct experiments on 4199 vehicle images of CompCars data set from the Chinese University of Hong Kong for vehicle front face detection test. The results demonstrate that the average coincidence degree of vehicle face is 73.74%, which is better than other object detection algorithms. In addition, the proposed method also has better generalization ability across data set without training.