流形群运动目标自动识别与跟踪模型结构及参数的最优配置

探讨了用流形群运动目标的形状、纹理特征，以及它们的空间面积的重叠度来构建多种适合流形群运动目标自动识别与跟踪的匹配模板的方法。通过最大欧几里得贴近度的择近原则，实现模板匹配，完成目标识别；通过对匹配模板的交替更新和交叉匹配算法，实现目标跟踪。为了提高识别与跟踪的准确度和效率，对识别与跟踪模型结构及参数进行了优化组合，建立了一种适合计算机自动识别和跟踪类似中尺度对流系统(MCSs)的流形群运动目标的优化模型，即多级串行和多级并行模板匹配的识别与跟踪模型，并提出了基于此模型的快速识别与跟踪算法及技巧。用优化了的多级串行识别与跟踪模型及快速跟踪算法，对1998、1999、2000、2002年6～8月的青藏高原上空MCSs进行了识别与跟踪试验。试验结果表明，其准确率高达90％。

A Robust Algorithm for Automated Identifying & Tacking Flow Shape Group Moving Targets

This paper discusses the technology of automatically identifying and tracking the flow shape group moving targets and puts forward several kinds of technology and methods for constructing matching templates. The templates matching has been realized by getting across the most of the Euclid pressing close degree choosing approximately principle,and the targets identifying has been achieved, and the targets tracking has been realized by updating elements of the matching templates by turns and using across matching arithmetic. To increase accuracy ratio of identifying and tracking targets, the model structure and parameters are optimized. Then the optimization model of tracing targets was established, which is based on multilevel-serial and multilevel-collateral matching templates. The arithmetic and skill of speediness identifying and tracking targets are also put forward with the model. By the optimization model and arithmetic, the experiments to identify and track MCSs in summer of 1998,1999 and 2000 over Tibetan Plateau are performed. The experiments results have shown that: this kind of the technology will very adapt to a computer to automatic identify and track these flow shape group moving targets such as the MCSs. And the optimization model and arithmetic are full of availability to identify and tracking the MCSs.