Building clustering method based on SOM neural network

Abstract

The invention provides a building clustering method based on an SOM neural network. The building clustering method based on an SOM neural network includes the steps: performing calculation of correlated characteristic parameters of buildings by means of the polygon layer data and the linear essential factor layer data of buildings; according to the calculated correlated characteristic parameters,by means of a Kohonen learning algorithm of the SOM neural network, performing preliminary clustering on a building group, and obtaining the preliminary building group clustering result; and using a line-rank scanning method to perform further fine clustering division on the preliminary clustering result, and obtaining the final clustering result of the building group. The building clustering method based on an SOM neural network realizes intelligent clustering of buildings, with map cognition capability, and can divide the adjacent buildings into one group under constraint of linear elementsin an urban area in which the building elements are dense and are distributed regularly. And in a suburb or country in which the buildings are distributed in a scattered manner, in the condition without road constraint, the building clustering method based on an SOM neural network can perform clustering according to the distance between buildings, and accords with the drawing requirement.

Description

technical field [0001] The invention relates to the technical field of computer map drawing, in particular to a building clustering method based on SOM neural network. Background technique [0002] Building features in large and medium scale maps are usually symbolized as discretely distributed polygonal graphics. As the scale of the map is reduced, features on the map become crowded and difficult to identify. Therefore, many buildings need to be deleted, merged, simplified, etc. Through the research on the synthesis process of artificial cartography, it is found that building synthesis includes two coherent steps: the first step is to divide the buildings into different groups, that is, building clusters; the second step is to target different building groups Groups perform different generalization operations. Automatic building synthesis is the application of computer programs to simulate these two steps. [0003] Regnauldls proposed methods for constructing and storin...

AI Technical Summary

Problems solved by technology

But existing data models or data structures cannot fully support intelligent cartographic generalization methods

[0010] (2) Cognitive theory has little guiding effect on automatic cartographic synthesis

[0012] So far, there are still many problems unresolved in the comprehensive research on the automatic mapping of residential elements, such as the comprehensive correlation between residential places and other elements is not enough, and the results of structured selection of residential places are not very good

Images