Share Email Print
cover

Proceedings Paper

An improved anomaly detection and classification algorithm of high-order statistics for hyperspectral images
Author(s): Li Lu; Wen Sheng; Xianzhi Zhang; Shihua Liu
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

An improved anomaly detection and classification algorithm based on high-order statistics is presented. In order to solve some challenging problems, such as initializing projection, quantifying of anomaly classes and evaluating the performances. Firstly, initialize the projection vectors used by the idea of global RX algorithm. It gives priority to the detection of the anomalies with powerful energy. Secondly, analyze the current data whether have anomaly information or not so that it determines the terminal conditions and the quantities of anomaly classes. Thirdly, use two methods to evaluate the classification performance quantitatively. One is to match the results in the condition of reference images to evaluate the effects of anomaly detection and background suppression, the other is to segment the resultant images to calculate some features such as the classification rate, the number of detected anomalies and the number of false alarms. Simulated and Experimental results show that the improved algorithm has the capability of robustness and better anomaly detection performances under complex unknown background than traditional algorithm does.

Paper Details

Date Published: 19 December 2013
PDF: 8 pages
Proc. SPIE 9045, 2013 International Conference on Optical Instruments and Technology: Optoelectronic Imaging and Processing Technology, 90451H (19 December 2013); doi: 10.1117/12.2037282
Show Author Affiliations
Li Lu, Air Force Early Warning Academy (China)
Wen Sheng, Air Force Early Warning Academy (China)
Xianzhi Zhang, Air Force Early Warning Academy (China)
Shihua Liu, Air Force Early Warning Academy (China)


Published in SPIE Proceedings Vol. 9045:
2013 International Conference on Optical Instruments and Technology: Optoelectronic Imaging and Processing Technology
Xinggang Lin; Jesse Zheng, Editor(s)

© SPIE. Terms of Use
Back to Top