Classification and characteristics of several common spatial indexes

here is an overview of the spatial index, click to download

2. 1 Grid IndexThe basic idea of grid index is to divide the study area into small grids according to certain rules, and record the geographic objects contained in each grid. When the user
When you make a spatial query, you first calculate the grid where the query object resides, and then quickly query the selected geographic objects through the grid. The grid index algorithm is broadly divided into
Three categories: Spatial index algorithm based on fixed grid partition, spatial index algorithm based on multi-level grid and adaptive hierarchical spatial indexing algorithm.
2. 1. 1 based on the spatial index algorithm based on fixed grid partitioning a map into a fixed mesh of a*b to index the map target falling into each grid,
So just retrieve the original area of I, mine B,
To achieve the purpose of fast retrieval. The advantage of this algorithm is that it is easy to operate and has certain adaptability when the amount of data involved is not big and complicated operation is not needed.
2. 1. 2 An empty-smelling index algorithm based on multi-level grids divides a map into small chunks of the same size that will fall into the small
The map object inside the block is stored in the corresponding storage area of the chunk, and the small blocks can be divided into smaller chunks to create multilevel indexes. The optimization of the algorithm
The point is that the retrieval is more efficient than a purely grid index, reducing the number of specific comparisons. But the fine degree of meshing is not guaranteed to be optimal. Right
The object at the edge of the mesh does not have a good solution, and does not take into account the effect of the horizontal and vertical distribution of the map target on the meshing.
2. 1. 3 Adaptive Hierarchical grid Spatial index algorithm
The grid size is determined by the bounding rectangle of each specific map object, which avoids the human factor of grid dividing in grid index. The advantage of the algorithm is that the grid
The partition stability is automatic, with the size of the external rectangle of the map target as the partition basis, avoids the duplicate storage, and has certain improvement in the storage efficiency. Insufficient is
is the algorithm implementation complex, before indexing, you must know the map of the outside of the target rectangle's length and width, sorted by its area size: After indexing, insert or delete
In addition to the operation, the map targets involved in the external rectangular area is not the size of the original area, you need to reorder the efficiency will decrease.
2. 24 Tree Index
The four-tree index, similar to the grid index, is also the grid partition of the geographical space, the geographical empty question recursive four points to construct the four-tree,
Until the self-setting termination condition (for example, the number of associated elements per node does not exceed
3, more than 3, then four points), eventually forming a hierarchical four-fork tree. Each leaf node stores the list of entity identities associated with this area and the
Domain geographic range, non-leaf nodes store only the geographic extent of the region. As the generation and maintenance of quadtree is relatively simple, and the data
When the image distribution is more homogeneous, the spatial query index based on Quadtree can obtain higher efficiency of space data inserting and querying.
2. 3 R-Tree family index
This is +. An index algorithm for object-oriented segmentation, which divides the empty objects by scope, each node corresponds to a region and a disk page, and a non-page node
A disk page that stores the area range of its nodes, and the leaf node's disk page stores the bounding rectangle of all the empty objects in its region.
2. 3. I algorithm of spatial index based on R-Tree
R-Tree algorithm is a hierarchical data structure Dynamic index algorithm, which is a natural extension of B-tree in k-dimensional space, and is a highly balanced tree. The R-Tree consists of three nodes, the root node, the middle node and the leaf node,
The Ask node represents a rectangle in the dataset's empty question, which contains the smallest bounding rectangle for all other child nodes, and the leaf node stores the bounding rectangle of the actual object.
The R-tree allows the nodes to cover each other, which allows the R-tree to maintain a high space-to-ask utilization and maintain the tree balance. No need to predict compared to the mesh arch L
The index range of the entire study area can establish an empty-question index, which reduces the storage redundancy of large geographic objects. Because it organizes the index structure by data, it is a
A fully dynamic index structure that does not require periodic index reorganization. However, too much overlap between nodes does not guarantee the uniqueness of the search path, and some even
Retrieves the entire tree, resulting in reduced query efficiency.
2. 3. Spatial index algorithm of 2 skeleton r+ tree
Sellis proposed R ten trees in 1987. R ten trees are similar to R-trees, except that there is no overlap between the r+ of the sibling nodes in the tree, thus eliminating
In addition to the dead regions that are generated by allowing nodes to overlap when the R tree divides space, the number of invalid queries is reduced and the efficiency of the spatial index is increased, but the insert delete operation
Reduced efficiency. All rectangles of the middle node of the r+ tree are disjoint. If an object's MBR is split by a rectangle in two or more IH tree high-level nodes,
Each item associated with a rectangle in these non-leaf nodes has a subsequent leaf node that points to the object. This increases the height of the tree, but the performance of the search operation can be greatly improved.
2. 3. 3 Spatial index algorithm based on r+ tree
Beckmann raised the r+ tree in 1990. R. The place where the tree is optimized relative to the R-Tree is the force re-insertion algorithm, and the idea is: when the new empty question is
When the insertion of an index entry causes the node to overflow, select nodes in the same layer to be adjusted to postpone the node splitting, so as to optimize the whole knot of R-Tree
The purpose of the construction. The spatial index algorithm based on R. Tree improves the space utilization and reduces the number of node splitting, but also increases the computational cost of CPU.
2. 4 Pyramid Index
Berchtold proposed the Pyramid method in 1998, which is based on a special unbalanced segmentation strategy which optimizes the high-dimensional data, and the principle is to first
D Virin Q is divided into 2d pyramids, the center point of the shared data space is the vertex, and then each pyramid is divided into a data page parallel to the pyramid base. Gold
The index structure of the tower is to transform the high-dimensional data into one-dimensional data and operate with the B + tree. The advantage of the pyramid index structure is that when the scope query is processed, the performance of this index structure is suboptimal for other index structures,
And the efficiency of query processing will not decrease with the increase of dimension, because it is more possible to make full use of the points in the pyramid question in the B + tree when processing the fan-week query.
The fact that the tree is on the same data page. However, the advantages of the pyramid index structure are based on the uniform data distribution and the hypercube query, for those coverage
The query of the boundary is not ideal, and the data in the real world is seldom subjected to uniform distribution.
3 Conclusion
In these indexes, different indexes have their own advantages and disadvantages as well as the scope of application. You need to choose which kind of empty-ask index method, according to the actual situation and
needs to be determined. In practical application, it is also a method of using multiple index structures to complement each other. Efficient method of empty-question indexing a ' straight is a lot of scholars and experts to study
Subject, there are many problems that need to be entered. ' Steps to solve such as improvements in efficient indexing tree algorithms, optimization of complex null query methods, and query insertion operations
optimization of the algorithm.