Development of Image Recognition Method using Dynamic Knowledge Model for Knowledge Discovery and Extraction

提出了一种无先验信息的动态知识模型，用于从现实世界的大数据中发现和提取动态知识。通过改变动态知识模型中的知识结构，在不需要外部数据的情况下自主地提高知识发现和提取的准确性。利用动态知识模型，提出了一种不需要先验知识的图像识别方法。该方法通过扩展游戏树搜索来实现知识发现过程。在实验中，我们开发了一种从两幅运动图像中检测目标的目标检测系统。

A dynamic knowledge model without prior information for the discovery and extraction of a dynamic knowledge from large data on real world was proposed. Knowledge structures in the dynamic knowledge model are changed to increase the accuracy autonomously in the discovery and extraction with no data from the outside. An image recognition method without prior knowledge about image was developed using the dynamic knowledge model. Process of knowledge discovery in the method was implemented by extending the game tree search. For experiments, we have developed a target detection system detecting a target from two moving images.