A multispectral remote sensing image unsupervised change detection method based on information expansion

Abstract

The invention relates to a multispectral remote sensing image unsupervised change detection method based on information expansion. The method comprises the following steps: 1) waveband spectral information expansion generated based on a nonlinear waveband: performing waveband spectral information expansion on any two original wavebands of multispectral images of a previous time phase and a next time phase respectively to obtain a newly-expanded nonlinear artificial spectral waveband; 2) wave band spatial information expansion based on multi-scale morphological reconstruction: respectively carrying out spatial reconstruction operation on the original wave band and the nonlinear artificial spectrum wave band to obtain a spatial expansion wave band feature set, and carrying out differential operation to obtain a differential wave band set; 3) Unsupervised multi-class change detection. Compared with the prior art, the method has the advantages that limited original wave band expansion andspatial information reconstruction are achieved, and automatic and unsupervised detection of multi-class changes is achieved.

Description

technical field [0001] The invention relates to the field of phase multispectral remote sensing image processing, in particular to a method for non-supervised change detection of multispectral remote sensing images based on information expansion. Background technique [0002] The dynamic changes of the earth's surface reveal the evolution process of the environment from the time dimension. Continuous earth observation by earth observation satellites can detect the impact of human activities and natural factors on the environment and land cover, which is of great significance to the study of global change. Change detection technology is one of the most important technologies in earth observation research. It can automatically or semi-automatically identify the actual occurrence of land cover / land cover between two or more images in the same geographical area in different satellite observation time periods. Take advantage of change. Among the many sensors carried on satellit...

AI Technical Summary

Problems solved by technology

[0003] Conventionally, when using multi-temporal and multi-spectral images for change detection applications, due to the limitations of the data source itself, such as the poor quality of image bands (including uncalibrated water vapor bands, image bad sectors, band noise, etc.), the space between bands Location inconsistency, inconsistency between multi-temporal data (such as cloud, shadow and other interference), and the limitation of the spectral range of the band (such as only covering the visible light band, no near-infrared, mid-wave infrared), etc., and then using very few bands When performing multi-type change detection, the original spectral band cannot fully reflect the real ground change information, which increases the difficulty of change detection

Although methods based on kernel functions or deep neural networks can provide a way to solve such problems to a certain extent, they all require a long-term tuning process and a huge amount of computation, and the large number of training samples required by these methods is changing in practice. detection applications are difficult to obtain

Images