The traditional method of detecting vehicle location by induction coil vehicle has a lot of problems, such as error in judgment, inconvenient of replacing the equipment and high cost. To solve these problems, this paper proposes a binocular vision method to detect vehicle location. This method adapts to all kinds of light changes. Selecting mode of day or night is determined by the different intensity of light surrounding environment. The mainly works of mode day include the median filtering of difference image, the extraction of binary image and the real-time update of partitioned background. Based on this, the paper proposes a method to detect vehicle location which combine the background subtraction and frame difference methods. This method can accurately extract the threshold value to acquire precise outline of the object. While in night mode, the infrared lamp is opened for switch to infrared CCD acquisition function. It's according to the values of RGB which in infrared lamp irradiation area to detect the vehicle location. This method has solved the problem of vehicle lights interference to video acquisition. Finally, the method is experimentally verified, and the results show that it can accurately detect the location of the vehicle and control the brake lever ups and downs.