A novel shape recognition method is presented.The proposed approach is based on the fact that the distribution of geometric features of a shape is invariant to rotation and scaling.An even-grid-polar mapping is applied to the shape and two invariant projections(one is up to circle-shift) are performed in the sampling plane.Thus

two extreme point vectors of the projections are collected.Then

corresponding weight vector features are calculated based on extreme point's style and scope to reduce the instability of the projections caused by geometrical deformation.The position of each element in the weighted vector illustrates the distribution of local parts with higher/lower pixel density in a shape and the magnitude is related to the pixel density.After analyzing the range of variations of the weight vectors caused by geometrical deformation

special templates and special matching scheme based on reparametrization are exploited so as to make this method invariant to rotation

scaling

translation (RST) and small non-rigid deformation.Experimental results show that the proposed method is valid.