Object detection device

An object detection device receiving a pair of brightness images as front view images in front of an own vehicle every predetermined detection period, transmitted from a stereo camera, the object detection device comprising a computer system including a central processing unit, the computer system being configured to provide:

• a distance image generation part generating a distance image, which represents a distribution of a distance value corresponding to a position on the brightness images on the basis of the pair of brightness images;
• a road surface estimation part estimating a road surface on the basis of the distance image generated by the distance image generation part;
• an object detection part which groups positions, each of the positions having a distance value located on/over the road surface estimated by the road surface estimation part and satisfying a predetermined relationship, and detecting an object on the basis of the grouped positions;
• and
• an abnormality detection part detecting occurrence of abnormality in the brightness image or the distance image obtained during a current detection period based on comparing at least one of the brightness image and the distance image with an image obtained during a previous detection period which is before the current detection period,
• wherein
• the road surface estimation part estimates the road surface at the current detection period on the basis of the road surface estimated at the previous detection period based on the abnormality detection part detecting the occurrence of abnormality in the brightness image or the distance image obtained at the current detection period.

An object detection method performed by a computer system including a CPU as a central processing unit and a computer-readable memory mounted on an own vehicle equipped with a stereo camera, wherein the CPU receives a pair of brightness images, as front view images of the own vehicle on a road on which the own vehicle is driving, transmitted from the stereo camera every predetermined detection period, and the CPU stores the pair of brightness images into the computer-readable memory,

• the object detection method comprising steps of:
• generating a distance image, which represents a distribution of a distance value corresponding to a position on the brightness images on the basis of the pair of brightness images, and storing the distance image into the computer-readable memory;
• estimating a road surface state of the road on the basis of the distance image, and storing the road surface state of the road into the computer-readable memory;
• grouping positions as grouped positions, each of the positions having a distance value located on/over the road surface state of the road and satisfying a predetermined relationship, and detecting an object on the basis of the grouped positions;
• and
• comparing at least one of the pair of brightness images and the distance image stored in the computer-readable memory with a corresponding brightness image or distance image stored in the computer-readable memory,
• obtained during a previous detection period which is before a current detection period,
• detecting occurrence of abnormality in the brightness image or the distance image obtained during the current detection period on the basis of a comparison result, and
• estimating the road surface state of the road is estimated at the current detection period on the basis of the road surface of the road estimated at the previous detection period based on the occurrence of abnormality in the brightness image or the distance image obtained at the current detection period being detected.