A change detection method for remotely sensed images using a statistical test for change recognition and change pattern discrimination

✍ Scribed by Hiroshi Okumura; Hiroshi Hanaizumi; Toshinori Yoshikawa

• Publisher: John Wiley and Sons
• Year: 1999
• Tongue: English
• Weight: 887 KB
• Volume: 30
• Category: Article
• ISSN: 0882-1666
• DOI: 10.1002/(sici)1520-684x(19991115)30:12<34::aid-scj4>3.0.co;2-z

✦ Synopsis

This paper proposes a new method for detection of changed regions and for classification of their change pattern, DETECT, from two remotely sensed multispectral multitemporal images acquired at the same location and at different times. The method applies a local regression model to a rectangular region of interest based on a changed pixel intensity distribution pattern between two different times. A statistical test is examined to judge whether the differences of regression coefficients are significant or not. Then, these regions are subdivided into subregions when their regression coefficients show a significant difference. Compared with the conventional method, the proposed method is free from impulse noise. Also, it is not influenced by apparent changes caused by sensitivity degradation in sensors or by the location changes of the light source. The threshold, which will directly affect the result of change recognition, is determined theoretically. Therefore, the proposed method yields inarbitrary results. These features are examined by simulations using a set of artificial images. The method is demonstrated to be successful when applied to actual satellite images.