Polarization SAR terrain classification method based on deep learning and distance metric learning

Abstract

The invention discloses a polarization SAR terrain classification method based on deep learning and distance metric learning. The polarization SAR terrain classification method comprises the realization steps that (1) images are inputted; (2) filtering is performed; (3) features are extracted; (4) training samples and test samples are selected; (5) a stacked sparse auto-encoder is trained so that the deep features of a training sample set and the deep features of a test sample set are obtained; (6) a distance metric learning classifier is trained so that a classification result is obtained; (7) the classification result is colored; and (8) the colored classification result graph is outputted. The images are classified by using the polarization SAR terrain classification method based on deep learning and distance metric learning so that feature extraction is relatively comprehensive and reasonable, the classification result is more consistent with real terrains, time complexity is reduced and classification precision is enhanced.

Description

technical field [0001] The present invention belongs to the technical field of image processing, and further relates to a PolarimetricSyntheticApertureRadar (PolSAR) feature classification method based on deep learning and metric learning in the technical field of image classification. The invention can be used for feature extraction and object classification of polarimetric SAR images. Background technique [0002] Polarimetric SAR image classification is an important step in the image interpretation process, and it is also an important research direction of polarimetric SAR image processing. Polarization SAR can obtain richer surface feature information than traditional single-polarization SAR. In the face of these large-scale and complex data, it is unrealistic for traditional methods to quickly process them and achieve high classification accuracy. , so it is urgent to propose some classification methods that can handle large data and have low time complexity. [0003]...

AI Technical Summary

Problems solved by technology

However, the disadvantages of this method are, first, for the face database it uses, the classifier is directly trained with training samples, and the image cannot be represented hierarchically

This method first performs Cloude decomposition on each pixel of the image, and then divides the image initially according to the entropy H and scattering angle α obtained from the decomposition, and then performs K-wishart iteration on the division results, which can improve the classification accuracy to a certain extent , to reduce computational complexity, but the shortcomings of this method are that, firstly, each pixel is decomposed by Cloude to obtain two features of entropy H and scattering angle α, and the extraction of features is not comprehensive and reasonable enough, which makes the present invention It is not close enough to the original data, so the classification accuracy is limited

Second, for data with a larger amount of information, the processing speed is too slow and the time complexity is too high

Images