| 摘要 |
In the present disclosures, after elimination processing (filtering) ahead of time, subject candidates extracted from video images are classified. Here, in said elimination processing, subject candidates are presumed to have a predetermined size set ahead of time for desired categories of subject candidates, and on the basis of the size of the subject candidates within a current image, the distances (ZO) from a vehicle to the subject candidates are calculated, the distance (”Z) of vehicle travel from the point in time a past image was captured until the point in time the current image was captured is calculated, the size of the subject candidate in the current image is reduced by a change ratio that is on the basis of the distance traveled and the distance to said subject candidate, and the size of the subject candidate in the past image and the size (Ghv) of the subject candidate reduced by the change ratio are compared. Subject candidates of which the difference between the two is determined to be above a predetermined value are eliminated. Because detailed classification is performed after having coarsely eliminated what is not a desired subject, computation load can be eased and mis-determinations can be reduced. |
