Tensor collaboration graph discriminant analysis remote sensing image feature extraction method

Abstract

The invention discloses a tensor collaborative graph discriminant analysis remote sensing image feature extraction method, and aims to provide a supervised feature extraction method which is low in complexity and good in feature extraction performance. According to the technical scheme, the method comprises the steps of intercepting a three-dimensional tensor data block with each pixel as the center; dividing experimental data into a training set and a test set in proportion; calculating the Euclidean distance between a current training pixel and each type of training data, and constructing adiagonal weight constraint matrix; secondly, designing an L2 norm cooperative representation model with constraints, and constructing a graph weight matrix and a tensor locality preserving projectionmodel; solving a projection matrix of each dimension of the corresponding tensor data block; and finally, obtaining a training set and a test set of three-dimensional low-dimensional representation byusing the low-dimensional projection matrix, expanding the training set and the test set into a column vector form according to feature dimensions, inputting the extracted low-dimensional features into a support vector machine classifier for classification, judging the category of the test set, and evaluating the feature extraction performance according to a classification effect.

Description

technical field [0001] The invention relates to image feature extraction in the field of image processing, in particular to a graph discriminant analysis feature extraction technology for remote sensing images, in particular to a tensor collaborative graph discriminant analysis remote sensing image feature extraction method. Background technique [0002] In many application fields, especially in cloud computing, mobile Internet, and big data applications, a large amount of high-dimensional and high-order data will be generated, and the mathematical form of tensor can be used to properly represent these data with multi-dimensional structures. These data often contain a large amount of redundant information, which needs to be effectively reduced in dimension. In pattern recognition, feature extraction (dimension reduction) and classification are two key steps. Most of the classic feature extraction and classification algorithms are based on vector data, which needs to be vect...

AI Technical Summary

Problems solved by technology

[0008] Aiming at the problems of high spectral dimension of hyperspectral data, large information redundancy, high complexity of existing methods, and insufficient spatial information mining, the present invention proposes a supervised feature extraction method with low complexity and good feature extraction performance. Make up for the shortcomings of existing feature extraction methods

Images