A fully polarized SAR image classification method based on intensity statistical sparsity

Abstract

The invention discloses a fully polarimetric SAR image classification method based on sparse strength statistics and is applied to fully polarimetric SAR image classification and target recognition. The method includes 1, inputting scattering intensity matrix; 2, acquiring SAR image statistics features; 3, acquiring SAR image sparse features; 4, training a classifier for classification; 5, optimizing initial classification results; 6, outputting classification results. According to the method, the scattering intensity information of three fully polarimetric SAR image channels is utilized, the feature information of space among the three channels is extracted, the specific target of the fully polarimetric SAR image can be classified effectively, and the specific features of the specific target can be retained completely.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to a full-polarization synthetic aperture radar (Synthetic Aperture Radar, SAR) image classification method based on intensity statistics sparseness in the technical field of target recognition. The invention can be applied to extract statistical features and sparse features of scattering intensity values ​​in full-polarization synthetic aperture radar SAR images, and accurately classify specific targets. Background technique [0002] As a representative of microwave remote sensing technology, high-resolution omnipolar synthetic aperture radar contains richer backscattering information of targets, which is an inevitable development trend in the field of SAR. The understanding and interpretation of full-polarization SAR images involves many disciplines such as machine learning, pattern recognition, signal processing, fuzzy logic, etc., and belongs to the category of im...

AI Technical Summary

Problems solved by technology

However, the disadvantage of this method is that this method only extracts the information of one channel of the full-polarization synthetic aperture radar SAR image, and does not involve the information between the three channels, which will inevitably cause the loss of classification information and classification results. Inaccurate, the maximum likelihood estimation method is applied in the Markov random field MRF algorithm, which increases the computational complexity and lengthens the time to obtain the result

However, the disadvantage of this method is that since the number of sub-apertures is inversely proportional to the ground spatial resolution when decomposing the image, the ground spatial resolution is reduced due to the increase of the number of sub-apertures, making the final classification accuracy restricted

Images