Research On Semantic Expression Of Traffic Event Based On Ontology And Description Logic

The semantic recognition and expression of traffic events with the ontology and description logic is a research hot spot of the traffic field. Howerer, the existing computer image semantic analysis technology can not acquire and express automatic the high-level semantics of traffic events based on the traffic scene video image sequences. It is because of lacking for the object recognition method and "the semantic gap problems" between low-level image visual concept mapping to the event’s high-level semantics. Most traditional recognition and expression methods of traffic events focus on the studying of pure image data for analysis instead of semantic analysis. So it is very meaningful to research recognition and expression methods of the high-level semantics of traffic events.This paper is based on provincial funded projects "road safety reasoning based on traffic situation assessment" and the research consists of three parts: the research on recognizing the objects from concept hierarchy, the research on mapping low-level concepts to high-level semantics and the development of recognition and expression systems of the high-level semantics of traffic events. The main research contents are as follows:(1) The attributes of traffic object in the image, for example, color, texture etc, can be extracted by using the improved "LST-KDE" algorithm, Camshift algorithm based on HSI color space and the Hough transform; then, the Conceptual Hierarchy Model of Traffic Image is modeling to map the objects with different attributes to the concepts by appling the ontology; finally, a method of establishing traffic domain ontology automatically is put forward to realize the recognition of traffic scene objects from concept hierarchy and provide a conceptual basis for the following semantic reasoning.(2) In order to solve the problem of "semantic gap", this paper uses description logic as the basic of logical expression and reasoning to use logical expression based on predicate logic and Tableau algorithm to build the bridge between concept of low-level and high-level semantics. Then, this paper proposed the method of using low-level concept to descript high-level semantic expression.In this method, description logic character set is the key of concept map to semantics and semantic reasoning. In order to describ the spatial topological relations and direction relations between scene objects, this paper introduce RCC8 space topological and conical space direction model.(3)This paper presents a traffic incident high-level semantic recognition, expression a nd reasoning system, including attribute fusion, semantic mapping and event identification and expression.Firstly, use the best sample image sequence with good training to divid the complete traffic sitiuation evolution process into several sub-events. Then, use FSM auto mata and description logic RQL query language to ratiocinate and identify each sub-event. In teh end, the advanced semantic identification and expression of the complete traffic eve nts can be realized.The innovation of this research: A structured semantic expression system has been built by using formalization natural language(ontology and description logic), which can be recognized by the computer; Traffic incident situational evolution process in the scene can be indentified and ratiocinated by analysising the semantic of the traffic scene image based on simulating human brain thinking. These invovations can provide new ideas and methods for traffic scene understanding, image retrieval and semantics recognition and expression of traffic events.