| 摘要 |
The invention relates to the field of image processing applied to the detection of targets and, more precisely, to the field of the estimation of the ground plane in images. It relates to a method of determining a triplet of parameters defining a ground plane on the basis of a depth image. The method of determining the ground plane comprises in particular a step of determining several triplets of parameters each defining a ground plane, a step of spatial filtering of the parameters and a step of temporal filtering of the parameters. According to the invention, the temporal filtering of the parameters is dependent on an indicator R3d representative of the quality of the data of the depth image which are available for the determination of the ground plane. The temporal filtering of the parameters may also be dependent on an indicator Rpl representative of the quality of the ground plane determined after spatial filtering. Likewise, the spatial filtering may also be parameterized as a function of the indicators R3d and Rpl. Globally, the invention implements the idea according to which the more the plane estimated for a depth image makes it possible to "explain" the points of this image, the more accurately the planes of the subsequent images may be determined. In particular, for the spatial filtering, this involves a ground plane search in a more limited space and, for the temporal filtering, a less significant consideration of the ground planes estimated previously. |
