A SAR image sample block selection method based on the aggregation characteristics of sketched line segments

Abstract

The invention discloses a SAR image sample block selection method based on the aggregation characteristics of sketch line segments, which mainly solves the problem of excessive redundant samples of one sliding window sample set in the pixel subspace segmentation of the mixed aggregation structure of SAR images in the prior art. The implementation scheme is as follows: extracting sketch map according to the sketch model of SAR image; regionalizing the sketch map to get the region map, and obtaining the extremely heterogeneous region of SAR image according to the region map; extracting all sketch line segments with two-sided aggregation characteristics in each aggregation region, and solving a set of rectangular blocks formed by the sketch line segments and the sketch line segments of K nearest neighbors; according to the coordinates of the rectangular blocks in the rectangular block set, obtaining a sample block set by sampling and expanding in the corresponding extremely inhomogeneousregion. The method can construct the sample set of the extremely heterogeneous region of the SAR image, and the constructed sample set can more fully represent the structural characteristics of the extremely heterogeneous region.

Description

technical field [0001] The invention belongs to the technical field of image processing, in particular to a SAR image sample block selection method based on the aggregation characteristics of sketch line segments, which is applied to the training sample set of highly inhomogeneous areas that are not connected to each other in the mixed pixel subspace of a synthetic aperture radar SAR image Determined, and further used for feature learning of extremely heterogeneous regions in SAR images. Background technique [0002] Synthetic aperture radar (SAR) is an important progress in the field of remote sensing technology, which is used to obtain high-resolution images of the earth's surface. Compared with other types of imaging technologies, SAR is especially suitable for large-area landmark imaging. It can penetrate clouds, rain and snow, and has the ability to work around the clock. With the rapid development of SAR image technology, more and more SAR image data are obtained, and...

AI Technical Summary

Problems solved by technology

The disadvantage of this method is that when learning the structural features of the disconnected areas in the aggregation area, the sample blocks sent to the deconvolution learning network are not selected, which makes the sample size large, resulting in network training time Long, and the space complexity is high, making the cost of network learning features larger

The disadvantage of this method is that when performing feature learning on highly heterogeneous regions that are not connected to each other in the pixel subspace of mixed aggregation structure features, this method is to collect every extremely heterogeneous region with a sliding window Random selection of the sample blocks obtained by the machine learning model makes the features learned by the machine learning model random, that is, it may be necessary to select random samples multiple times and learn features multiple times before it is possible to learn the extreme Good feature representation in heterogeneous regions affects the accuracy of pixel subspace segmentation of SAR images with mixed aggregation structures and increases training costs

Images