An image size changing section obtains a size changed image by changing the size of one of two original images captured by a pair of cameras. If an edge of an object has many oblique components, the edge is difficult to detect as a vertical edge but, when the image is horizontally reduced, an oblique edge becomes close to a vertical edge. For this reason, feature end points are extracted with reliability by a feature extracting section and, thereby, an object is recognized and the distance to the object is determined reliably.