Projection Pursuit Hyperspectral Image Segmentation Method Based on Transfer Learning

Abstract

The invention discloses a migration learning-based projection pursuit hyperspectral image segmentation method, which belongs to the field of image processing technology, and its technical gist is: using the multi-band characteristics of hyperspectral image data, each band is regarded as a Gray-scale images are analyzed and studied for each image; using the similar but different characteristics of images in different bands, transfer learning is introduced into the projection pursuit clustering algorithm; the labels of the source domain image data are obtained through the ground object marker map , using the known label knowledge to guide the unlabeled image data of the target domain to obtain the optimal projection direction and optimal subspace, thereby improving the segmentation accuracy. The invention has the advantages of utilizing the existing prior knowledge and improving the segmentation precision, and can be used in military reconnaissance means and civil and industrial fields.

Description

technical field [0001] The invention belongs to the technical field of image processing, relates to hyperspectral image segmentation, and can be used in military reconnaissance means and civil and industrial fields. Background technique [0002] Hyperspectral remote sensing is one of the most important developments in the field of remote sensing since the 1980s. It has become a hot topic in the field of international remote sensing technology research in the 1990s, and it will also be the frontier technology of remote sensing in the next few decades. Hyperspectral remote sensing technology uses imaging spectrometers to image surface objects simultaneously with tens or hundreds of bands with nanoscale spectral resolution, and can obtain continuous spectral information of surface objects, realizing spatial information, radiation information, and spectral information of surface objects. The synchronous acquisition of , has the feature of "map integration". Hyperspectral images...

AI Technical Summary

Problems solved by technology

In the traditional hyperspectral image segmentation, the K-means clustering algorithm and other clustering algorithms are used, but the K-means clustering algorithm is used to directly segment a single image or a multi-band image, and its segmentation accuracy is often not high. The clustering algorithm directly segments a single image or a multi-band image. Although its segmentation accuracy is improved compared with the K-means clustering algorithm, it cannot make full use of the existing prior knowledge, and its segmentation accuracy is still not high.

Images