Hyperspectral image abnormal point detection method based on selective kernel principal component analysis

Abstract

The present invention provides a detecting method for the unusual point of the high spectrum image based on the selective nuclear main composition analyzing, and relates to the high spectrum image analyzing and defecting technology field. The present invention aims at solving the problem that the prior high spectrum image defecting method cannot conduct character extracting on the unusual point effectively, and produce more false alarm, and cannot detect the unusual point effectively under high background disturbance. The present invention has the procedures that the data is normalized, and the nuclear main composition is analyzed; a sliding window is constructed in the main sub-quantity; the third moment and the fourth moment of the pixel in the sliding window are calculated and compared the moments with a preset value; the value is recorded after reading all over the main sub-quantity; all the main sub-quantity are processed; the maximal main sub-quantity is selected; the unusual point detecting on the selected main sub-quantity is conducted by the RX operator, and the detecting result is output. The present invention can extract and select the character of the unusual point object in the high spectrum image effectively, and can lower the false alarm rate, and can detect the unusual point normally under high background disturbance.

Description

technical field [0001] The invention relates to the technical field of hyperspectral image analysis and detection. Background technique [0002] Abnormal signals are also called mutation signals. Abnormal points and irregular mutations in the signal often carry very important information, which is one of the important characteristics of the signal. In image analysis and processing, an outlier is usually defined as a part that is different from the background category in a small neighborhood, or a part that undergoes a mutation, which is defined with reference to a different background model. Outliers often only occupy a small number of pixels in the image, or even exist within one pixel. The background model is determined by selecting a large-scale area from the image or the reference data of the local neighborhood near the test pixel, and is mainly divided into a global background model and a local background model. The main difference in the definition of global and loca...

AI Technical Summary

Problems solved by technology

[0011] The purpose of the present invention is to solve the problem that the existing hyperspectral image detection technology cannot efficiently extract and select the featur

Images