Extraction of visual descriptors is a crucial problem for stateof- the-art visual information analysis. In this paper, we present a knowledgebased approach for detection of visual objects in video sequences, extraction of visual descriptors and matching with pre-defined objects. The proposed approach models objects through their visual descriptors defined in MPEG7. It first extracts moving regions using an efficient active contours technique. It then computes visual descriptions of the moving regions including color, motion and shape features that are invariant to affine transformations. The extracted features are matched to a-priori knowledge about the objects¢ descriptions, using appropriately defined matching functions. Results are presented which illustrate the theoretical developments.