| 摘要 |
<p>A method for the detection of a target present in at least two images of the same scene acquired simultaneously by different cameras comprises, under development conditions, a prior target-learning step, said learning step including a step of modeling of the data X corresponding to an area of interest in the images by a distribution law P such that P(X)=P(X2d,X3d,XT)=P(X2d)P(X3d)P(XT) where X2d are the luminance data in the area of interest, X3d are the depth data in the area of interest, and XT are the movement data in the area of interest. The method also comprises, under operating conditions, a simultaneous step of classification of objects present in the images, the target being regarded as detected when an object is classified as being one of the targets learnt during the learning step. Application: monitoring, assistance and security on the basis of stereoscopic images.</p> |
