Automatic Semantic Annotation Method For IoT Sensory Data Research And Implementation

The continuous development of the Internet of Things(IoT),Cyber-Physical Systems(CPS),and sensor technologies have facilitated the application of a wide variety of IoT devices,resulting in a huge amount of IoT sensory data.These sensory data with different sources,having different data formats and meanings,can not achieve semantic interoperability among IoT services and applications.Therefore,to realize information interaction among IoT entities,the meaning of heterogeneous data needs to be described uniformly.Semantic annotation techniques can realize the unified description of data and interoperability among IoT entities.However,the manual and semi-automatic annotation methods available in the current study still require a large amount of expert knowledge and experience and suffer from insufficient annotation capability and inefficiency.In addition,the existing automatic semantic annotation methods are more focused on the applications in the Internet text domain,and there are fewer studies on automatic semantic annotation methods for IoT sensing data.To address the above problems,this thesis introduces clustering analysis techniques and ontology development and rule inference techniques in the Semantic Web into automatic semantic annotation methods for IoT-aware data.The main research work of the thesis is as follows.1.To address the problem that the semantic information implied in the data cannot be extracted due to the heterogeneous characteristics of multiple sources of IoT sensing data.The machine learning clustering technique is studied for the analysis and processing of IoT sensing data to obtain the clustering centers and the number of categories of the data and to complete the extraction of the implicit semantic information in the data.In this thesis,we design a k-centroid-based Fast Search for Density Peaks(KCFSDP)clustering analysis method by combining the advantages of the automatic determination of cluster centers by the clustering by fast search and find of density peaks(CFSDP)clustering algorithm and the stability of the k-centroid clustering algorithm,which is suitable for processing large data sets,and complete the classification of perceptual data and extraction of semantic information.2.To address the problem that the semantic sensor setwork(SSN)ontology of the world wide web consortium(W3C)organization is insufficient for the description of IoT-aware data.The ontology modeling guidelines,evaluation criteria,and modeling methods are studied,an ontology for the IoT data domain is designed according to the characteristics of IoT data,and the ontology modeling tool Top Braid Composer is used to build this ontology to semantically describe the information extracted from IoT-aware data.3.To address the problem that the manually constructed ontology of the IoT data domain can not be updated automatically.Study the rule inference based on semantic web rule language(SWRL),design the automatic updating method of ontology based on SWRL rules using inference based on-premises and results,and reason the newly obtained semantic knowledge to realize the dynamic updating of ontology concepts.4.Combining the KCFSDP clustering method,IoT data domain ontology,and designed SWRL rules,we propose an automatic semantic annotation method for IoT-aware data,and use resource description framework(RDF)to semantically describe the data information to achieve automatic semantic annotation for IoT-aware data.5.Develop an automatic semantic annotation system for IoT sensing data,use the laboratory hardware and software resources,and then build a validation platform by combining related technologies and validating its functional modules.According to the experimental results,the automatic semantic annotation method proposed in this thesis is feasible.Through comparison experiments,the research method in this thesis has a greater efficiency improvement compared with the existing semantic annotation methods.