摘要
本文提出了一种模式识别理论的新模型 ,它是基于“认识”事物而不是基于“区分”事物为目的 .与传统以“最佳划分”为目标的统计模式识别相比 ,它更接近于人类“认识”事物的特性 ,故称为“仿生模式识别” .它的数学方法在于研究特征空间中样本集合的拓扑性质 ,故亦称作“拓扑模式识别” .“拓扑模式识别”的理论基点在于它确认了特征空间中同类样本的连续性 (不能分裂成两个彼此不邻接的部分 )特性 .文中用“仿生模式识别”理论及其“高维空间复杂几何形体覆盖神经网络”识别方法 ,对地平面刚体目标全方位识别问题作了实验 .对各种形状相像的动物及车辆模型作全方位 880 0次识别 ,结果正确识别率为 99 75 % ,错误识别率与拒识率分别为 0与 0 2 5 % .
A new model of pattern recognition principles was proposed, based on matter cognition instead of matter classification in traditional statistical pattern recognition. This model is closer to the function of human being than traditional statistical pattern recognition using optimal separation as main principle. So this model is called the bionic pattern recognition, while its mathematical basis is topological analysis of sample set in the high dimensional feature space, therefore it is also called the topological pattern recognition. The basic idea of this model is based on the continuity in the feature space of similar kinds of samples. Experiments on recognition of omnidirectional rigid objects on the same level were carried out with this model using neural network. Covering the high dimensional geometrical distribution of sample set in the feature space. Many animal and vehicle models (even with similar shapes) were recognized omnidirectionally in 8800 tests, showing that the correct recognition rate is 99.75%, while the error rate and the rejection rate are 0 and 0.25% respectively.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2002年第10期1417-1420,共4页
Acta Electronica Sinica
基金
国家自然科学基金项目 (No 60 1 350 1 0 )
