Classification-oriented hyperspectral remote sensing image noise band detection method

Abstract

The invention provides a classification-oriented hyperspectral remote sensing image noise band detection method. The method includes: adopting a visual interpretation method to obtain a sample of a certain type of pure ground object in a hyperspectral remote sensing image; acquiring spectral information of the samples, wherein all the samples and the corresponding spectral information form a training sample set with the spectral information, and the spectral information corresponds to different spectral bands; performing training through the training sample set to obtain a plurality of decision trees, and aggregating the plurality of decision trees to form a random forest; calculating the importance of a spectral band in the random forest; determining an importance threshold P of the spectral band; and calculating the importance of Q spectral bands in a certain hyperspectral remote sensing image which is actually input, and automatically detecting a severe noise band according to the importance threshold P of the spectral bands. The method has the beneficial effects that the serious noise wave band in the hyperspectral remote sensing image is detected, and support is provided for improving the classification precision.

Description

technical field [0001] The invention relates to the field of hyperspectral remote sensing image processing, in particular to a classification-oriented hyperspectral remote sensing image noise band detection method. Background technique [0002] Hyperspectral remote sensing combines imaging and spectral technology to simultaneously obtain spatial-dimensional image information and spectral-dimensional spectral information of an area of ​​interest, presenting an image cube structure, which has the characteristics of "map-spectrum integration". Compared with multi-spectral remote sensing images and high-spatial-resolution remote sensing images, each pixel records spectral information of dozens or even hundreds of continuous bands, and often records spectral information between 400 and 2500 nm with a spectral resolution below 10 nm. The characteristics of ground objects provide spectral features that can distinguish the physical properties of different substances, and can detect ...

AI Technical Summary

Problems solved by technology

However, at present, the classification accuracy of most classification algorithms is affected by the noise band, and the corresponding classification interpretation accuracy is often reduced for severe noise bands. Therefore, it is necessary to carry out research on hyperspectral noise band detection methods for classification tasks.

Images