Double-path convolutional neural network remote sensing classification method taking spatial neighborhood relationship into account

Abstract

The invention discloses a double-path convolutional neural network remote sensing classification method taking a spatial neighborhood relationship into account. The double-path convolutional neural network remote sensing classification method comprises the steps of: data acquisition; data preprocessing; thematic map production; multi-scale segmentation; visual interpretation; first-order neighborhood acquisition; experimental sample set selection; patch size selection; training data generation; mat format conversion; convolutional neural network structure construction; subject single-path neural network training; first-order neighborhood single-path neural network training; complementarity analysis; and double-path convolutional neural network training taking the spatial neighborhood relationship into account. The invention aims to provide a method for regarding a subject and a first-order neighborhood of the subject as input image data on the same scale, realizing the enhancement of sample information through increasing neighborhood information of samples and reducing interference caused by other types of samples, so as to realize the improvement of classification precision and provide a reference for classification decision-making.

Description

[0001] technology neighborhood [0002] The invention relates to the field of remote sensing image classification, in particular to a two-way convolutional neural network remote sensing classification method in consideration of spatial neighborhood relations. Background technique [0003] Remote sensing image classification is currently a hot research topic in remote sensing technology, and how to accurately and efficiently classify remote sensing images has always been an important research content in the field of remote sensing. In recent years, with the development of artificial intelligence technology, neural network has gradually become an effective classification and processing method for remote sensing images. Compared with the traditional statistical classification methods, the neural network has the characteristics of learning ability and fault tolerance, and does not need to make assumptions about the probability model, so it is suitable for dealing with various pro...

AI Technical Summary

Problems solved by technology

[0008] The technical problem to be solved by the present invention is to provide a two-way convolutional neural network remote sensing classification method that takes into account the spatial neighborhood relationship in view of the defect of low classification accuracy in the prior art

Images