贝叶斯专家系统分类器中专家知识的自动提取

Realization of Knowledge Auto-extraction for Bayesian Expert System Classifier

专家知识提取是长期以来遥感专家系统分类器应用过程中存在的瓶颈问题,该文重点解决如何实现贝叶斯专家系统分类器中专家知识自动提取和知识库建立的问题。基于参考样点统计分析得出的规律,提出专家知识与参考信息间关系的假设,并设计专家知识自动提取方法,即分类类型先验概率和条件概率估计方法。为了验证专家知识自动提取方法的准确性和有效性,应用模拟数据和实际研究区域数据进行分类并评价其精度。结果表明,基于参考样点统计分析能够获得较高精度的各类型分布先验概率和条件概率,从而实现专家知识自动提取,有效解决现有贝叶斯专家系统分类器中存在的瓶颈问题。

Knowledge acquirement has been a long--term bottleneck in the application of expert system classifier in remote sensing. This paper focused on solving the question of expert knowledge auto-- extraction and knowledge-- base construction for Bayesian expert classifier. Based on the statistics characteristics of sample points, hypothesis between knowledge and reference information and auto--extraction method was set up. The conditional--probability for each class was realized based on statistical analysis of reference samples. In order to validate the accuracy and validity of the method, classification and accuracy assessment on both simulated data and real study--site data had been carried out. Results showed that the conditional--probabilities could be effectively estimated based on the statistical analysis of the reference samples, which means that the knowledge could be extracted automatically for the Bayesian expert system classifier. The landscape classification using expert system classifier with the auto--extracted knowledge showed a higher accuracy than the classifications of maximum likelihood classifier or decision tree learning classifier. This study solved successfully the bottleneck of knowledge acquirement in Bayesian expert system classifier.