Several main data models of geographic information system

1, topological relation data model

Topological relation data model organizes and stores geometric features based on topological relations, which is characterized by centering on topological connections among points, lines and surfaces, and their coordinate storage is dependent. The main advantages of this model are compact data structure and clear topological relationship. The topological relations pre-stored in the system can effectively improve the efficiency of the system in topology query and network analysis, but there are also some shortcomings:

The operation efficiency of a single geographical entity is not high. Because on the one hand, the topological data model is oriented to the whole spatial area, emphasizing the relationship between geometric features, on the other hand, it does not pay enough attention to the fact that geographical entities with complete and independent significance exist as individuals. Therefore, when adding, deleting and modifying a geographical entity, a series of files and relational database tables will be involved, which will not only complicate the program management, but also reduce the execution efficiency of the system.

It is difficult to express complex geographical entities. Due to the requirements of topological relationship organization, in the topological relationship model, a complete simple entity sometimes needs to be decomposed into multiple geometric features (for example, a highway is a complete entity, but in order to record its topological adjacency information, a complete entity can only be decomposed into multiple geometric features if it is divided where it is adjacent to other highway entities. All entities are treated in this way, so we say that the topological data model faces the whole area and undivided geometric features, not the geographical entities in the eyes of users. Complex geographical entities are composed of many simple entities, and naturally they are often decomposed. The overall organizational characteristics of topological data model determine that it cannot effectively express this organic aggregate composed of many independent entities.

It is difficult to realize fast query and complex spatial analysis. In the topological data model, geographical entities are decomposed into points, lines and basic geometric features, which are stored in different files and relational tables, so operations, queries and analysis involving independent geographical entities will take up more CPU time, especially in complex large-area spatial analysis.

It is difficult to update locally and maintain and expand the system. Because the data organization and storage of geographical space are based on the basic geometric features (points, arcs and surfaces), the complex topological relations stored in the system are the data foundation of GIS work. When some local entities change, the topological relationship of the whole layer will have to be rebuilt. Such a system will affect the whole body, and it needs more energy to maintain and expand, and it is easy to make mistakes.

It is worth noting that the topological data model can also be realized in an object-oriented way, but the object at this time is a geometric feature, which is not divided by other elements, and is often a part of an independent geographical entity, rather than a complete independent geographical entity. This is one of the important manifestations of the essential differences between the topological relational data model and the entity-oriented data model in the next section.

2. Entity-oriented data model

It is called "entity-oriented" to emphasize that this data model takes a single spatial geographic entity as the basic unit of data organization and storage.

Contrary to the above topological model, this model takes independent, complete and geographically significant entities as the basic unit to express geographical space. In the concrete organization and storage, the coordinate data and attribute data of entities (such as local topology, which are also stored in tables) can be stored in file system and relational database respectively, or in relational database (coordinate data and attribute data can be stored in the same table or in two tables). The storage mode of SDE ESRI Company is divided into four tables, and a layer table and a spatial index table are added. The layer table is located on the server side and used for layer management and maintenance. The spatial index table (server-side) adopts grid index, which is used for quick search of entities).

The entity-oriented data model is realized by a completely object-oriented software development method. Each object (independent geographical entity) not only has its own independent attributes (including coordinate data), but also has its own behavior (operation), and can complete some operations by itself. Although the entity-oriented data model can be organized according to the topological relationship, the model mentioned by the author here emphasizes that there is no dependency between the coordinate storage of objects (especially the coordinate storage of surfaces and lines), which is the essential difference between it and the topological relationship model. This model can well overcome several shortcomings of topological relational data model, and has the advantages of easy entity management and modification, easy query and spatial analysis, and more importantly, it can easily construct any complex geographical entity that users need. This model conforms to people's thinking habits of looking at the objective world and is convenient for users to understand and accept. At the same time, the entity-oriented data model naturally has the advantage of convenient system maintenance and expansion.

This model is the latest data model adopted by popular GIS software, but it also has some shortcomings:

Topological relationships need to be built temporarily. Because the entity-oriented data model is centered on geographical entities, it does not organize, store geographical entities and express geographical space based on topological relations, so topological relations do not exist at the beginning, but are temporarily derived when needed, which requires certain system resources. Perhaps there is a view that when organizing data by entities, topological relations can be saved in the attribute table of entities from the beginning, and topological relations are not necessarily temporary. However, after careful analysis, it can be found that this scheme is not feasible for entities composed of multiple geometric features (such as rivers composed of different components) because the topological relationship cannot be recorded effectively and accurately. In fact, this scheme is only suitable for entities composed of a geometric feature, but it is still a topological data model in essence, and its shortcomings are exactly the same as those described in section 2. 1 above, so it is not a real entity-oriented data model.

Dynamic segmentation and inefficiency of network analysis. In the node-arc-polygon topological relationship chain, there are four explicit topological tables: node-arc table, arc-node table, arc-polygon table and polygon-arc table. With these four relational tables, we can directly find the topological properties of any node, arc segment and polygon, which is convenient for dynamic segmentation, network analysis and other topological analysis related to topological relations. GIS based on topological data model can do this easily. However, because storing all four topology tables will double the space cost of the system, some softwares only store two of them (such as the early version of System 9) or combine arc-node and arc-polygon tables into one (the version before Arc/Info 8.0), the implicit table can be exported from the table that shows the existence. Even so, GIS based on topological data model still has high efficiency in the query and analysis of topological relations, while entity-oriented data model will naturally reduce the efficiency of topological query and analysis because it needs to temporarily construct topological relations as needed. Of course, the constructed topological relationships can be stored for future use.

Duplicate common points and edges between storage entities. Because the entity-oriented data model is based on geographical entities for data organization and spatial expression, each geographical entity is completely stored (stored at the level of points), and each object is independently stored when storing coordinates, so it will inevitably lead to repeated storage of * * * points and * * * edges between entities.

It is difficult to manage, analyze and process at the level of geometric features. Geometric features are simple figures, such as points, arcs and polygons. Sometimes, the attributes of geometrical features, which constitute the same entity, are very different (for example, the areas of the sides that constitute the land parcel are different, and the types of roads that constitute the traffic closed loop are also different, and so on). ), so it needs to be processed at the next level of geographical entities-geometric feature level, and the topological data model can be processed directly, while the entity-oriented data model needs to locate and decompose related geographical entities first, which reduces the performance of the system in this respect. In essence, it is not difficult to draw a conclusion that this model thinks that the attributes of geometric features that make up the same entity are the same, so it ignores the attribute differences between geometric features, which makes it difficult to locate the storage and processing mechanism of the system at the level of geometric features.

It is difficult to realize cross-layer topology query and analysis. If we put this problem into the topological relationship model, it will be easier to solve, because the adjacent elements of each element already exist in advance, and they are not only hierarchical, but also have actual geographical attributes, so we just need to track the adjacent elements and get their geographical attributes. For the entity-oriented data model, it can't be effectively solved, because when the topological relationship is generated temporarily, geometrical features generally belongs to the same layer, so it can't automatically generate cross-layer geographical attributes, and it can only be solved after further processing. Obviously, this method is inefficient.