SAR image classification method based on SAR-SIFT and DBN

Abstract

The invention discloses a SAR image classification method based on SAR-SIFT and DBN. The problem of synthetic aperture radar image classification is mainly solved. The classification process comprises the following steps: (1) inputting an integer type SAR image matrix; (2) transforming the SAR image matrix; (3) extracting SAR-SIFT features; (4) max pooling; (5) normalization; (6) training four restricted Boltzmann machines RBM; (7) training a softmax classifier; (8) building a deep belief network DBN; (9) classification; and (10) calculating the classification accuracy. According to the invention, the SAR-SIFT features of synthetic aperture radar images are extracted, the deep belief network DBN is adopted, the features are learned layer by layer, the information integrity of the radar images is retained, depth information is mined, and a good classification effect is achieved. The method can be used for SAR image classification.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to a synthetic aperture radar-based scale invariant feature transform (Synthetic Aperture Radar-Scale invariant feature transform, SAR-SIFT) and a deep belief network (Deep Brief Network) in the technical field of image classification , DBN) synthetic aperture radar (Synthetic Aperture Radar, SAR) image classification method. The invention can be applied to target classification and recognition of SAR images. Background technique [0002] Synthetic Aperture Radar (SAR) can work around the clock and around the clock, and the image resolution it obtains is comparable to that of optical images. The classification of SAR images is a key branch in the field of synthetic aperture radar imaging. When using SAR images to classify ground objects, feature extraction and selection are crucial. Typical features in SAR images include peaks, textures, regions, edges, intensities...

AI Technical Summary

Problems solved by technology

However, the disadvantage of this method is that it is required to learn a dictionary for each type of training data and obtain a synthetic dictionary, and it depends on the dynamic integration selection of each atom in the synthetic dictionary, which leads to the training time of the classification method. Long, computationally complex, and not robust, reducing the average accuracy of multiple classification runs

This method uses the data information of the image, the correlation information between the image and the relevant information of the image on multiple scales, and extracts important information from a small amount of weak information, that is, learns SAR from the training data marked with keywords. The image classification model can greatly reduce the difficulty of obtaining accurate training data, and overcome some local uncertainties in SAR image classification, but the disadvantage of this method is that multiple correlations between images at the same time In the process of information processing, the information between data is cut too fast, and a small amount of weak information used for training loses many important details, resulting in low accuracy and insufficient robustness of the classification method.

Images