作者: Jiann Shu Lee , Yung Nien Sun , Wen Huei Lin , Chin Hsing Chen
DOI:
关键词:
摘要: SUMMARY A method to recognize planar objects undergoing affine transformation is proposed in this paper. The based upon wavelet multiscale features and Hopfield neural networks. feature vector consists of the transformed extremal evolution. evolution contains information contour primitives a manner, which can be used discriminate dominant points, hence good initial state network obtained. Such initiation enables converge more efficiently. normalization scheme was applied make our scale invariant reduce distortion resulting from normalizing object contours. employed as global processing mechanism for matching made suitable whose shape arising an transformation. improved guarantee unique stable results. new evaluation scheme, computationally efficient, evaluate goodness matching. Two sets images, noiseless noisy industrial tools, were test performance method. Experimental results showed that not only effective robust under