节点文献
一种神经网络体视协同算法
A COMBINATORIAL STEREO VISION ALGORITHM IMPLEMENTED AS NEURAL NETS
【摘要】 描述一种多分辨率体视协同算法.该算法用二分正交小波作多分辨率分解,采取边匹配边重建,二者互相修正协同进行的策略,并以神经网络主要是Hopfield网络为其实现形式.试验结果表明,该算法是合理和有效的.
【Abstract】 Described here is a combinatorial multiresolution stereo vision algorithm. In the algorithm, dyadic orthogonal wavelet is used for multiresolution decomposition, and, a matching while reconstruction strategy is taken, in which matching and surface reconstruction correct each other and proceed cooperatively. The algorithm is implemented in the form of neural nets, mainly, Hopfield net. Results of primary tests show its rationality and effectiveness.
【关键词】 体视;
特征匹配;
表面重建;
多分辨率分析;
神经网络;
【Key words】 Stereo vision; feature matching; surface reconstruction; multiresolution analysis; neural network.;
【Key words】 Stereo vision; feature matching; surface reconstruction; multiresolution analysis; neural network.;
【基金】 航空科学基金
- 【文献出处】 自动化学报 , 编辑部邮箱 ,1998年03期
- 【分类号】TP18
- 【被引频次】5
- 【下载频次】58