Robot visual planar locating method based on improved RBF network

For robot visual location, a locating method based on non-linear mapping planar orientation using improved radial basis function (RBF) network is presented. To improve the adaptability on the camera´s pose change, this method uses the pitch degree of camera as one of the input of RBF network, makes use of the non-linear approach of RBF network fits the non-linear mapping relationship from the image plane to the special plane, and positions the targets. In order to reduce the interaction between the RBF´s feeling domains, improve the locating accuracy, a Gaussian core function which could inhibit the nuclear feeling domain is presented, and the improved RBF network take the Gaussian core function for the use of RBF, the locating accuracy is enhanced. In order to reduce the redundancy in RBF network´s initial center selection, an improved method for initial center selection is proposed, this method form the samples in accordance with the norm, taking into account the infection of network´s implicit nodes and preset a redundant threshold. Setting different pitch degrees, the results of locating experiments express this method is effective in the application of target locating on the special plane. The comparison of results of locating experiments shows that, under the same condition, the precision of this method is higher than the precision of the visual locating methods based on planar homography matrix estimation which is in common use.