Research On Scene Similarity Retrieval Based On F-Histogram

The information technology revolution is sweeping the global world,with the rapid growing of spatial data,how to effectively use the spatial data and to retrieve the required information quickly in the massive spatial data is the bottleneck of current spatial data management.Current spatial data retrieval and query methods are lagging behind.Because of the characteristics of spatial data,its retrieval has the characteristics of fuzzy description,imprecision and non-standard.Traditional spatial data retrieval methods have the following disadvantages: they are different from people's subjective cognition and cannot express the user's query intention exactly.At the same time,there is no detailed description of the spatial characteristics of spatial data.These shortcomings restrict the accuracy and efficiency of the traditional search methods.In order to eliminate the inconvenience and limitation of traditional retrieval methods,improve the utilization of spatial data,give full play to its spatial characteristics,and make users directly express the retrieval needs in spatial scene retrieval.In this paper,we study in depth of the spatial similarity search and propose a novel approach for spatial similarity retrieval.The research work of this paper is as follows:(1)Based on the existing spatial relationship description model,we analyze the advantages of F-histogram in describing spatial orientation relations,and expound the deficiency of F-histogram in describing topological relations.Firstly,a model of orientation relation similarity calculation based on F-histogram is proposed.Secondly,based on the concept of constraint relaxation of spatial topological relations,analytic hierarchy process and conceptual neighborhood graph of the region-region relations on 9IM,a weight-based constrained relaxation model of topological relation is proposed.Finally,we combine the above two models,and propose a scene similarity measurement model based on the orientation relationship and weight-based topological relationship.So that it can meet the scene search needs and make the calculation result in line with people's recognition.(2)Due to the limitation of common spatial scene search strategy,we improve the memetic algorithm according to the characteristics of spatial scene retrieval.Firstly,the query map and the query sketch and the reference geographic information system(GIS)database are represented as attribution relational graphs(ARG),and the scene matching problem istransformed into the matching problem of subgraph-matching problem.Then,a population initialization algorithm,evolutionary operation,and three local search algorithms for scene retrieval is proposed.The experimental results show that the method of scene similarity retrieval based on F-histogram has better retrieval accuracy.