A target segmentation algorithm based on multivariate statistics and RX anomaly detection

Qiandong Guo; Bing Zhang; Lianru Gao; Xu Sun; Wenjuan Zhang

doi:10.1117/12.918399

24 May 2012 A target segmentation algorithm based on multivariate statistics and RX anomaly detection

Qiandong Guo, Bing Zhang, Lianru Gao, Xu Sun, Wenjuan Zhang

Proceedings Volume 8390, Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XVIII; 83902N (2012) https://doi.org/10.1117/12.918399
Event: SPIE Defense, Security, and Sensing, 2012, Baltimore, Maryland, United States

Abstract

Anomaly detection of hyperspectral image is active topic in the field of remote sensing image processing. Reed-X Detector (RXD) algorithm developed by Reed and Yu is a Constant False Alarm Rate (CFAR) anomaly detection method founded on multivariate statistical analysis theory, as the same form with Mahalanobis distance. RX detector could enable researchers to exploit targets that people particularly want from their surroundings according to the spectral distinct. So RXD is practicable in real scenes, and then becomes a focus in the field of target detection. RX detector has two common forms, Global-RX and Local-RX. They have different samples to estimate mean vector and covariance matrix. PCA is a common preprocessing step for dimension reduction. Interestingly, because it can also remove noises, performance could be improved by using principle components instead of all data. In addition, people often assume that RX result values submit the chi-square distribution, which often leads to an unacceptable high false alarm rate in setting χ²_α,p as threshold. So, how to get threshold value has been a difficult problem. This paper proposes a method based on multivariate statistical probability theory which can segment targets from image automatically. Instead of a constant threshold value, this segmentation target approach use an initial threshold calculated by RX result value histogram to separate backgrounds and targets samples, then calculate every pixel's posterior probabilities of background or target by assuming they all submit multi-dimensional normal distribution. Generally, the higher probability is considerable. The proposed method has been tested using AVIRIS data and the experimental results reveal that segmentation target approach has higher detection probability and lower false alarm rate compared with the traditional manual thresholding way.

Citation Download Citation

Qiandong Guo, Bing Zhang, Lianru Gao, Xu Sun, and Wenjuan Zhang "A target segmentation algorithm based on multivariate statistics and RX anomaly detection", Proc. SPIE 8390, Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XVIII, 83902N (24 May 2012); https://doi.org/10.1117/12.918399

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available

Members: $17.00

Non-members: $21.00 ADD TO CART

PROCEEDINGS
7 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Target detection

Detection and tracking algorithms

Image segmentation

Sensors

Statistical analysis

Hyperspectral target detection

Principal component analysis

Show All Keywords

Keywords/Phrases

Search In:

Publication Years