An object recognition apparatus 10 includes a candidate image extraction unit 12 which extracts a candidate image part 22 from a pickup image 21, a distance calculation unit 13 which calculates the distance of the candidate image part 22, and a candidate image determination unit 14 which determines that, in the case where a predetermined k number or more of the candidate image parts 22, the real space positions of which belong to a determination area R having the distance thereof lying within the range of a second predetermined distance exceeding a first predetermined distance and the width thereof being a predetermined width or less, are extracted, the candidate image parts 22 which belong to the determination area R are less likely to be the image parts of a pedestrian.