A high spectrum exception detection method purifying backgrounds based on the local density

Abstract

The invention provides a high spectrum exception detection method purifying backgrounds based on the local density. The method comprises the steps of: firstly acquiring an initial background corresponding to a currently detected pixel by using a concentric double-layer window model; then calculating the local density of each pixel of the initial background; setting the maximum exception proportionand selecting the pixel corresponding to the minimum local density according to the proportion; dividing the background by using the maximum between-cluster variance method; detecting a high spectrumimage based on an LRXD exception detection method; setting a threshold value, marking pixels with the detection values greater than the threshold value as exceptional points. The method removes exceptional data in backgrounds by purifying the backgrounds, thereby facilitating analysis of differences in targets and backgrounds and effectively reducing the false alarm rate.

Description

technical field [0001] The invention relates to a hyperspectral anomaly detection method, in particular to a hyperspectral anomaly detection method based on a local density purification background. Background technique [0002] Hyperspectral image is a new type of three-dimensional remote sensing data that combines map and spectrum with high spectral resolution and many continuous spectral bands. This provides rich discriminative information for object detection. According to whether the object detection utilizes prior information (generally the object spectrum), it can be divided into two categories: supervised object detection and unsupervised object detection. Unsupervised target detection is also called abnormal target detection. In the case of unknown prior information, anomaly detection directly uses the spectral difference between the background and the target to detect abnormal points in the image. Anomaly detection methods provide a solution to the detection and r...

AI Technical Summary

Problems solved by technology

However, in actual situations, hyperspectral images generally have a variety of ground objects, and are affected by shadows, light, and atmospheric interference. The background of the acquired hyperspectral images is complex and changeable.

Therefore, the multidimensional Gaussian distribution model cannot fully reflect the distribution characteristics of real hyperspectral image features.

As long as there is 0.5% data pollution, there will be a problem of covariance distortion

For hyperspectral images, the complexity of the background distorts the statistical information, which makes the false alarm rate of the RX detection method higher

Images