A three-dimensional object detection device has a camera, a three-dimensional object detection unit, a lens cleaning device, a lens state assessment unit and a controller. The camera has a lens for forming an image of an area rearward of a vehicle. The three-dimensional object detection unit detects a three-dimensional object rearward of the vehicle based on the captured images. The lens cleaning device sprays cleaning fluid to clean the lens of the camera. The lens state assessment unit accesses whether the lens is in a predetermined state subject to control based on a timing at which cleaning fluid is sprayed on the lens. The controller suppresses detection of the three-dimensional object by retaining detection or assessment results for a predetermined length of time that were obtained immediately before the lens was assessed to be in the state subject to control, upon assessment that the lens state is subject to control.