摘要 |
PROBLEM TO BE SOLVED: To accurately and efficiently recognize an object that passes an observation position. SOLUTION: Two cameras pick up the image of a prescribed observation position on a road, after a characteristic extracting part 8 extracts edge constructing points on the image and associates between images of the edge constructing points, a three-dimensional measuring part 9 calculates a three- dimensional coordinate that corresponds to each edge constructing point and further, a projection processing part 10 projects each calculated three- dimensional coordinate point on a virtual flat plane that is vertically located along the longitudinal direction of the road. A matching processing part 11 executes matching processing with a two-dimensional model of various vehicles about the projection result. As a result, when a projecting point that corresponds to any of the two dimensional models is recognized, the two-dimensional model and a matching position are separately stored in a model storing part 12 and a matching position storing part 13 and they are used for matching processing for the next inputted image. |