Quality Evaluation For Land Use Data Generalization

We adhere to the "quality win" and "delicate work" principle in almost every field. As is known to all of us, quality plays an important role in spatial data representation and applications. The quality problem results from the process of data acquisition, data processing, data analysis and data application. Map generalization aims at simplifying data and has been claimed to be one of the most important sources of quality problem. The reason is that map generalization has a greater influence on data quality than the general data processing way, such as data acquisition, transformation of the reference system, etc. On one hand, map generalization needs to delete the details of data. On the other hand, the generalization result must maintain the original spatial distribution and the global information of study area. Therefore, we need to find a balance between quality loss and quality accuracy for a reasonable map generalization. For this reason, the research on quality assessment of the map generalization result has important implications. Compared with the general quality assessment of spatial data, map generalization has its unique features which highlight the objects to be assessed, reference datum, the source of quality problem and the evaluation model.The objects to be assessed for map generalization include the general spatial data which mainly represented by the topographic map and the thematic data considering the semantic feature. And land use data is a kind of thematic data and it can provide rich information and mainly describe the polygon feature. And there is a great difference in the rule, operator and result from the general map generalization, which contributes to its own uniqueness in the quality problem. The general map generalization mainly focuses on spatial characteristic. However, the quality assessment of land use generalization pays more attention to thematic attribution, such as land use type, administrative level and land ownership, etc. For the moment, there are many research findings on quality assessment of spatial characteristic for map generalization, but fewer researches with the characteristic of system thorough technical strategy and method on the quality assessment of thematic attribution. Following completion of the second state land investigation, the requirement of land use data for state department and other government agencies is increasing. What is needed now is to establish database with multi-level for map generalization. Consequently, how to assess the quality of land use generalization has become a crucial issue.The subject of this paper is researching the quality assessment of land use generalization. For the characteristic of scale transformation, we expand the research of quality problem generated by land use generalization on the basis of quality assessment for the general spatial data. Starting with the generalization constraint, this paper deeply analyzed the quality of land use data generalization and developed some evaluation model, algorithms and operators. They work together and provide quantitative evaluation results from micro, meso and macro levels. The main contents of the paper are depicted as follows:(1) On the basis of analyzing the quality problem of land use data generalization, we put forward a two-dimensional quality evaluation index that combined with level and constraint. Land use data is usually composed of polygonal features. And it is characterized with complete coverage, seamlessness, little overlap and hierarchy in semantic. In addition, the constraint, operator and algorithm of land use data generalization are usually different from those of the general map generalization. Thus, there exists particularity in the quality problem of land use data generalization. Specifically, the quality problems appear in the inconformity of topology, area, semantic, land classification structure and spatial location between the original data and the generalization result. Based on the constraints, this paper provided the quality evaluation indexes of the generalization result for land use data by three kinds of geometrical shapes. And it also put forward a two-dimensional quality evaluation index that combined with level and constraint.(2) Take polygonal feature, for instance. The evaluation indexes of the three levels are generalized into graphical evaluation, statistical evaluation and semantic evaluation. And there is a specific quantitative evaluation method for each index at the corresponding level.(3) This paper has studied two evaluation methods which can help us understand the maintenance of area balance and spatial distribution of the land classification structure. Land use data is closely related to regional characteristic. Based on this, the landscape index analysis method is improved to evaluate the maintenance of area balance, which took into account the structure similarity and pattern stability of land use classification structure with the spatial correlation analysis. Considering land use type has the property of spatial autocorrelation, this paper brought in the exploratory spatial data analysis and realized the evaluation of spatial distribution for the land use type quantitatively with the overall and local Moran’s I index.(4) We proposed four elements of semantic quality, which were accuracy, consistence, completeness and similarity. And each element has its own evaluation model. Specifically, model for accuracy is based on the semantic membership. Model for consistence is based on area. Model for completeness is based on the quantity of features and model for similarity is based on information entropy. It is known to all that the semantic information of land use data possesses the characteristic of hierarchy. Thus, we built a hybrid membership model which consists of hierarchy structure and order statistic. And further, we constructed the semantic membership matrix for the land use map derived from the second national land survey, which quantitatively evaluated the semantic quality from the three levels mentioned above.(5) This paper designed the concept frame for quality evaluation of land use data generalization based on the summary of characteristics of land use data generalization, the evaluation index and the evaluation method. And we broke the evaluation process into three steps which were description of feature characteristics, description of evaluation indexes and integration of evaluation results. What is more, this paper constructed the quality evaluation model at three levels. Simultaneously, the weight matrix which took into account the scale and regional characteristic was also designed. The example analysis on real data which came from the results of the DOMAP software verifies the validity of each evaluation index. The experiments proved that all indexes proposed in this paper can truly reflect the quality of the generalization results for land use data. And it can also reveal the changing features of the quality of generalization result with scales.The research of this paper has expanded quality problem of the general spatial data into the professional data processing. For the particularity of land use generalization and scale transformation, we have put forward the evaluation index, model and algorithm in three spatial levels. In addition, this study combining the application of actual data from the second national land investigation project, selected some evaluation index and the weight. At the same time, some quantitative evaluation methods were given and proved by experiment. Hence, this research turned out to be of great practical significance. Moreover, the research on evaluation of semantic quality and the statistic evaluation method based on spatial distribution pattern and spatial structure is an important supplement to the existing research on quality evaluation for spatial data.