Research On The Theory For Spatial Data Quality Inspection And Assessment

With the rapid development and popularity of the geographic information technology, the reliability and quality of spatial data is one of the most important precondition to ensure the correctness of the results of the decisions and applications based on spatial data. Therefore, the study of spatial data quality control is very significant for geographic information science, geographic industry and national geographic state survey or monitoring. Specially, spatial data quality inspection and assessment is one of the key methods to control spatial data quality. The study of spatial data quality inspection and assessment, which is very important for the assurance of the reality, reliability and authority of the geographic state survey and monitoring, has important practical significance, especially for the First General Survey of Geographic Condition of China (FGSGCC). Generally, the process of the spatial data quality inspection and assessment involves three key sub-processes:determining applicable spatial data quality model, selecting the measurements of the determined spatial data quality elements and executing a suited spatial data quality assessment method. However, there are some obstacles in the three sub-processes because of the diversity of spatial data and the complexity of the spatial uncertainty. In the sub-process of determining applicable spatial data quality model, most of the spatial data models are only applicable for some special type or scale of spatial data though some research results such as the spatial data quality model based assessment have been achieved, which led to the impossibility that the various spatial data sets are inspected and assessed in the same condition and platform and thus increased the cost of the inspection and assessment of spatial data. In the part of measuring the spatial data quality elements in the spatial data model, the measurements of position accuracy, which is the core of the spatial data quality, is always the hot spot in the study of spatial data quality. The error models for measuring the position uncertainty can be divided into three types according to the representation types of spatial data objects, which includes the error model based on point, line and polygon. Current error band models for line objects are mostly based on the error distribution of the points, which only considers the error of two end points of lines and ignores the error of the middle parts between the two end points. Thus, it is unreasonable to measure the position error of line objects and very difficult to propose the quality indicator for line objects. In the sub-process of validation of spatial data, it is a great challenge to keep the reference data sets synchronous with the spatial data and get enough high quality ground reference data.Therefore, the dissertation aims to study the general spatial data quality model, a scientific error band model and indicator for line objects based on the line error distribution and a new quality validation model and framework for spatial data based on crowdsourcing from the process of spatial data quality inspection and assessment. Based on these research results, a general spatial data quality inspection and assessment system, which can inspect and assess various spatial data and improve spatial data quality control of FGSCC, is designed and implemented. The details are listed as follows:(1) The definition and contents of spatial data quality model are analyzed through the concept of spatial data quality model. By the analysis and comparison of the definitions and descriptions for spatial data quality model and quality elements in both domestic and overseas standards, one important feature about the structure of spatial data quality model and quality elements is concluded:one spatial data quality model is composed of various quality elements, one quality element can be divided into many sub-elements and check items and one check item can be described by one or more quality indicators. It means that the spatial data quality model is described in a multilevel hierarchy. Based on this the concept of object tree is introduced and the general spatial data quality model based on the object tree in concept level is proposed. The dissertation defines the general spatial data quality model based on the object tree from three aspects, which include the object members of quality object tree, the relationship among the object members and the logical model of the quality model. The proposed quality model, which can solve the problems brought by the diversity of spatial data quality model and requirements has the advantages of both the object-oriented design and object tree, such as the uniqueness of the member objects, abstraction, encapsulation, transitivity, polymorphism and extensibility.(2) With respect to the spatial data assessment model, a new general quality assessment quality model based on the proposed quality model is built. Through the detailed analysis on the procedure of the spatial data quality assessment the procedure can be divided into six important components:the final spatial data quality assessment, the quality assessment for quality elements, the quality assessment for quality rules, error rate computation, the qualified condition and quality mark. According to this and the proposed general spatial data quality model based on object trees, the member objects and the relationships among them are defined and the logical hierarchy of the spatial data quality assessment model is provided.(3) To measure the position error of line objects, the error distribution of whole line is analyzed and a new standard deviation band theory is proposed. According to the process of the modelling of the standard deviation of random points, the standard deviation band based on the error distribution of line objects is analyzed and its analytic expression is derived. With respect to the probability that one line object lies in the standard derivation band, one approach, which regards the mean value of the upper and lower probability of the standard deviation as the probability that the line object lies in the standard deviation, is proposed. Moreover, a faster approximated computation method is proposed for practical application of the probability. The standard deviation band of polyline objects is proposed and the general deviation band of line and polyline objects is built on the basis of the standard deviation band of line objects. According to the analysis of the features of the quality indicators of point objects, a new quality indication which is the area of the band based on the line deviation band theory is proposed.(4) On the validation of spatial data, the method and framework using crowdsourcing data as the reference data are proposed. Firstly, the advantages (e.g. real-time, rich information and materials and low cost etc.) and possible problems (e.g. unreliable quality, heterogeneity and uneven distribution etc.) are analyzed. With respect to the defects of crowdsourcing data, the validation idea that crowdsourcing data and the target spatial data are viewed as the observation results for the real data and then an evaluation model is built to validate the data, is proposed. Moreover, the framework using crowdsourcing data to validate spatial data is built. To further explore the feasibility of the proposed framework, the dissertation take the validation of the global land cover data as an example and a latent class model on the basis of crowdsourcing data is proposed. To improve the validation results of the method, the latent class model analysis based on the binary classification is developed. Finally, a validation experiment using the global land cover data set GlobCover 2009 made by the European Space Agency is designed and performed and the results show that the method can produce some good evaluation for some specified land cover types though some problems like reliability of the results exist.(5) According to the proposed general spatial data quality and evaluation model, the general framework for spatial data inspection and assessment is proposed and the general spatial data inspection and assessment system is designed and developed. The proposed framework, which involved the data layer, core layer, support layer, execution layer and presentation layer, is some kind of logical implementation form the general spatial data quality and evaluation model based on object trees. Specially, the implemented system on the basis of the proposed framework and models has been widely used in FGSGCC that have resulted in more efficient data quality checking and data acceptance.