A fuzzy supervised classification method for multi-band remote sensing images based on non-equal weighted distance

Abstract

The invention discloses a fuzzy supervised classification method for a multiband remote sensing image based on non-equal weight distances. In the method disclosed by the invention, the non-equal weight distances of the cluster center of the remote sensing image and various land utilization types in different wavebands are redefined by adopting the manually extractive spectrum information of a training region based on a traditional FCM (File Compare Mask) classification method; and the degree of each pixel belonging to the different land utilization types is judged by utilizing the non-equal weight distances as an evaluation criterion, thereby realizing the fuzzy supervised classification of the remote sensing image. The method disclosed by the invention determines the cluster center and the distance value through the spectrum characteristics of the training region and obviously improves the classification efficiency and the classification accuracy.

Description

technical field [0001] The invention relates to the field of computer remote sensing image classification, in particular to a remote sensing image fuzzy supervised classification method. Background technique [0002] Remote sensing image classification has always been a hot issue in the field of remote sensing applications. [1] . The diversity and complexity of the real geographical world lead to the common occurrence of mixed pixels in remote sensing image classification [2] , and the traditional either-or hardening method is difficult to satisfy this uncertain and vague geographical status quo. The remote sensing image fuzzy classification method based on the uncertainty description of sample attributes can better express and deal with the unclear class attributes in remote sensing images, and has become a hot spot in the field of remote sensing image classification. [3] Many experiments have proved that the fuzzy classification method is more effective in expressing th...

AI Technical Summary

Problems solved by technology

[0003] As an unsupervised classification method that does not require human intervention but only determines the cluster centers and classification results based on self-iteration, it has the following shortcomings when it is directly used in remote sensing image analysis [7] : First of all, the unsupervised classification method of remote sensing images is to classify the categories directly according to the similarity measure between the individuals to be divided, and then manually identify the corresponding actual ground object types. The degree of automation is high, but it is difficult to guarantee the classification accuracy [8] ; Secondly, the multi-band characteristics of remote sensing images determine that the variable input must be multivariate, which directly leads to the low efficiency of the FCM classification method that achieves the best classification results by self-iteration; finally, because the potential sample structure information is unknown , it is often impossible to obtain the corresponding classification results according to the classification purpose, so the determination of the one-to-one correspondence between each category in the results and the real ground objects is one of the difficulties in the existing classic fuzzy C-means algorithm [1]

Images