结合粗糙集理论的发展及其在模式识别领域中的优势和不足，提出基于抗噪粗糙集的三维目标自动识别方法。在压扩式非均匀离散化编码和全程归一化处理的基础上，对动态层次聚类分类方法进行了改进，通过对偶然性事件和不相容事件加权概率处理，以可信度的形式将粗糙集的规则训练和抗噪性能结合起来，并提出基于相对最小类间距离的分层识别方法，实现了粗糙集基础上的规则训练与样式识别。通过对多种三维目标的识别仿真表明，该方法具有较好的抗噪性能、处理效率和识别效果。%After briefly reviewing the development of the rough set theory and its advantages and disadvantages on target recognition ,this paper proposes the method of 3D target recognition based on noise-robust rough set .On the whole ,the data of Condition Attribute Set is disposed by the unequali-ty compand expand disperse technique and normalization w hich is to change all the feathers to corre-sponding data form zero to one before training and recognizing ,and the dynamic layered cluster algo-rithm is improved accordingly .And then ,we work out the non statistical probability weighted meth-od to cope with the occasional examples and the incompatible examples especially ,so the training and recognizing in the rough set and the robustness is banded together by the reliability calculated during the training in the rules reduction steps .Simultaneously ,the hierarchical least relative distances be-tween clusters is brought forward for recognition by the improved rough set theory .At last ,different kinds of 3D targets pictures with different phases are tested in this paper ,the simulation result show that the conceive is with good robustness ,high efficiency and fine recognition effect .
展开▼
