With the widening application of GIS in more domain, Geographic Information Systems (GIS) has become the popular tools from the desk of a fewer geographic scientist. The better geo-information retrieval approach and spatial analysis technique is desired as GIS application expanding in more scientific and trade fields. Traditional spatial analysis based on exact match requires sound enough GIS database background knowledge that the GIS user have. How much geo-information and its data quality depend upon how much the GIS user acquaint with the GIS database. The more the GIS user comprehend the GIS database and GIS knowledge, the more detailed geo-information knowledge the user can get, and the more exact result that spatial analysis will produce. Unfortunately, not all the GIS users possess so enough knowledge that can help them achieve the spatial analysis result they expected. Geographic information science desiderates new methodology to improve geo-information retrieval and data analysis.

There are similarity and dissimilarity among geographic objects and geographic phenomena. The similarity is macroscopical and it gives people more impetus to recognize the physical world. Similarity science usually takes similarity phenomena and similarity system as research objects, and focuses on the similarity discipline and its application of natural world, human society, engineering technique and cognition thought, and it has complete theoretic science system including cognition similarity, similarity analysis, similarity application in practice, similarity creation. Similarity theory, which is one of bases of thinking science, aims at the study on the law of similarity movement, similarity correlation and similarity creation of natural world, society and thinking fields. Similarity science dwells on not only similarity phenomena, but also acquainting similarity and similarity system, exploring similarity source, expounding how to formulate similarity and its evolution, analyzing and measuring similarity, and research on similarity law and its application.

This thesis introduces similarity science theory and methodology into geographic spatial analysis, similarity appearance and computational method, brings forward similarity based spatial analysis theory and method, and then investigate key technology of similarity-based spatial analysis. A hierarchical qualitative spatial representation model that integrated with qualitative theme, qualitative shape, qualitative topological relationship, qualitative direction relationship, and qualitative distance relationship transform complicated qualitative spatial representation and spatial relationship to standard SQL queries statement and improves geographic information retrieval and its spatial representation accord with human spatial recognition ways. The multi-factor similarity evaluation model is improved so that the user’s interaction will be simplified, and this evaluation based on the improved model will be more rational. A wholly research on similarity based spatial analysis theory and key technology can be found in this thesis, which emphasis on spatial theme similarity, spatial objects similarity and spatial scene similarity. I investigate and develop one computer software system of similarity based spatial analysis, and discuss its application through the examples of practice on tobacco planting. The thesis’ original contribution includes:

Although it is couples years ago since human recognized similarity and exercise similarity self-consciously, it is almost the last twenty years when similarity science is developed as a science subject. There is few research on similarity in information system, the rather that fewer research on geographic information system based on similarity as the particularity compared with other information systems. This thesis bring similarity science theory into geographic spatial analysis, the scope covers similarity phenomena including spatial theme similarity, spatial objects similarity and spatial scene similarity, similarity of spatial analysis restraint condition, similarity in all period of spatial analysis, as well as their computation methods. To the best of my knowledge, there is not such a full-scale research in similar domain. It is so academic valuable that it brings new methodology to geographic spatial analysis research.

Qualitative spatial representation is the important part of qualitative GIS. In the former research, qualitative spatial theme, qualitative shape, qualitative topological relationship, qualitative directional relationship and qualitative distance relationship were represented by separated representation model and their calculation was done by respective model with high calculation complexity and not according to human spatial recognition habits. This thesis puts forward a hierarchical grid-based qualitative spatial representation and calculation model, which integrated with the five qualitative spatial representation ways mentioned above. The scalability of this model which can represent hierarchical and multi-measure character of human spatial recognition and representation has been verified by human spatial recognition test designed in this thesis. The problem of “sharp change” in former spatial representation model is solved in this thesis. And the transformation from complicated spatial relationship calculation to standard SQL query statement reduces the computation complexity, and improves query efficiency; as a result, the qualitative spatial representation model is more workable.

The system framworkof similarity-based spatial analysis has been built on the basis of research of spatial similarity and similarity of spatial analysis procedure. The work includes:

The spatial analysis needs a multi-variable similarity evaluation model which is according to human similarity judgment and similarity recognition habits, as it should evaluate integrative similarity of so much natural condition factors as well as economic and social condition factors. In this thesis, I build a rank-order based multi-variable similarity evaluation model implemented in spatial analysis. The model, which makes integrative similarity evaluation on the basis of the rank of each similarity variable, reduces and simplifies the users’ labor. The experimental example verifies the efficiency of this model.

The similarity comparison between null value and non-null value is inevitable, as the null value in database is inevitable. The computing value of existing algorithms dealing with null value is not according to the reality, as the dissimilarity is exaggerated. The improvement in this thesis which substitutes the null value with the mean value or the most probable value in comparison is more reasonable and has been validated in tobacco planting applications.

There is much difference between spatial analysis systems based on similarity and ones based on exact match. Spatial analysis systems bring similarity into all phase of spatial analysis. This thesis implements spatial analysis based on similarity in computer software engineering way by examples. The similarity-based spatial systems don’t require that the user must have enough knowledge background of geographic database, and its analysis way and result accord with human spatial recognition habit from coarse to fine, from qualitative to quantitative. Compared with traditional spatial analysis system, the spatial analysis system based similarity has predominance in higher analysis efficiency, more comprehensive analysis result, more relaxed constraint, and more applicable value.

Tobacco planting is a complex agriculture production involving multi-factor, so this thesis implement SQL-based and sketch-based spatial analysis based on similarity and comparing between two spatial scenes. Spatial analysis based on similarity provides tobacco planting a new analysis method and managing measure, and is a novel method to acquaint tobacco planting rules indeed.