Tham khảo tài liệu 'humanoid robots - new developments part 3', kỹ thuật - công nghệ, cơ khí - chế tạo máy phục vụ nhu cầu học tập, nghiên cứu và làm việc hiệu quả | 62 Humanoid Robots New Developments Ak 1 Ck 1 -1 c - i x - i T 10 m n X m n X f 1 m X ak x - Ak 1 T c 11 m n vv X m n m X rk 1 . z k 1 z ik T k 1 z jk 1 XT x T . -in Xm n x - amn - Amn c x - amn - Am n c m 12 All vectors are column vectors and m in 9 represents the weighted average with respect to the posterior probabilities of cluster m. The parameters bm n and means nm 1 are estimated as before. The conditional probability follows then from the joint PDF of the presence of an object on at the spatial location p with pose Ộ size and depth d given a set of contextual image measurements c bk G x nk Xk G C ik Ck z I - x _ bm nG x nm n Xm n G C nm n Cm n p x I On c ---M -------- -------------- b c m n Ckn m 1 Object detection and recognition requires the evaluation of this PDF at different locations in the parameter space. The mixture of gaussians is used to learn spatial distributions of objects from the spatial distribution of frequencies in an image. Figure 13 presents results for selection of the attentional focus for objects from the low-level cues given by the distribution of frequencies computed by wavelet decomposition. Some furniture objects were not moved such as the sofas while others were moved in different degrees the chair appeared in several positions during the experiment while the table and door suffered mild displacements. Still errors on the head gazing control added considerable location variability whenever a non-movable object was segmented and annotated. It demonstrates that given an holistic characterization of a scene by PCA on the image wavelet decomposition coefficients one can estimate the appropriate places whether objects often appear such as a chair in front of a table even if no chair is visible at the time - which also informs that regions in front of tables are good candidates to place a chair. Object occlusions by people are not relevant since local features are neglected favoring contextual ones. Back-propagation Neural Networks
