To rapidly detect a surrounding obstacle without requiring much computing.
An edge image extraction part 2 extracts an edge image at a fixed interval from an image taken by a camera 1, and an image computing part 3 carries out averaging processing on the extracted image and stores it in a frame memory 4. Inside a detection area set by an area setting part 5, a density projection part 6 performs density projection on the averaged edge image, and if a density projection value above a fixed threshold value continues for a fixed time or more, a vehicle determination part 7 determines that there is an obstacle, a vehicle especially, on the periphery. In association with the action of a direction indicator 8, an alarm output part 10 outputs an alarm.
NAGASAWA SHUNSUKE
Takaaki Yasumura
Takeshi Oshio