Clustering-classification-based FILTERSIM simulation method

Abstract

The invention, which relates to the technical field of image processing, provides a clustering-classification-based FILTERSIM simulation method, so that a technical problem of improving a simulation effect is solved. According to the method, K objects are selected randomly from a training image as initial mass centers of K clusters; all objects in the training image are distributed to clusters with the mass centers closest to objects; the mass centers of all clusters are calculated again and then all objects are distributed again, the above-mentioned operation is repeated continuously until the mass centers of all clusters do not change, and then the training image is divided into K training blocks, so that each training block has one cluster; and then all training blocks are simulated byusing a FILTERSIM simulation method. The method provided by the invention can be applied to fields of weather forecasting and resource exploration and the like.

Description

technical field [0001] The invention relates to an image processing technology, in particular to a technology of a FILTERSIM simulation method based on cluster classification. Background technique [0002] The MPS method (multi-point geostatistical method) is one of the mainstream methods for stochastic simulation of spatial data. The FILTERSIM simulation method is a filter-based MPS method, which is an important branch of the MPS method and can be used for stochastic simulation of continuous and discrete variables. [0003] When estimating spatial data variables, the FILTERSIM simulation method usually uses filters to classify each pattern of the training image, and then simulates the estimated area based on the pattern classification. Each mode is transformed into a filter score space, which can greatly reduce the dimensionality of training images. In the simulation process, the filter is used to obtain the "filter score" of the data event in the estimated area, and then...

AI Technical Summary

Problems solved by technology

exist figure 1 , there are the following problems: 1) There is a "sub-score space" that is empty (the square pointed by the black arrow); 2) a black solid point surrounded by an elliptical dotted frame is very far away from other solid points in the "sub-score space" to which it belongs It shows that the feature pattern of the training image corresponding to the solid point is quite different from the feature pattern corresponding to the "sub-score space" to which it belongs; 3) in figure 1 There are only two black solid points in the "sub-score space" in the lower right corner, and they are far apart, indicating that they are quite different and should be divided into other "sub-score spaces"

It can be seen that the data pattern classification method in the existing FILTERSIM simulation method obtains poor division results, which in turn affects the simulation effect

Images