Object-Oriented Random Forest Classification for Enteromorpha Prolifera Detection with SAR Images

• Published in: 2016 International Conference on Virtual Reality and Visualization (ICVRV) pp. 119 - 125
• Main Authors: Cui Xie, Junyu Dong, Fangfang Sun, Lei Bing
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.09.2016
• Subjects: Algae; enteromorpha prolifera detection; Image segmentation; Monitoring; object-oriented classification; random forest; Remote sensing; Satellites; Sea measurements; Synthetic aperture radar; synthetic aperture radar (SAR)
• DOI: 10.1109/ICVRV.2016.27

In this paper, a novel framework, called Object -- oriented random forest classification (OORFC), is proposed for Entheromorpha prolifera (E. prolifera) detection based on RADARSAT-2 Synthetic Aperture Radar (SAR) imagery obtained from 2011 to 2013 in the coastal areas in the Yellow Sea. Firstly, each SAR image is processed by multi-scale segmentation to generate patches with homogeneous attributes. Then a random forest classifier is trained based on the features of these patches. Finally, for a new coming SAR images, E. prolifera can be detected by the generated random forest classifier and the operational monitoring of E. prolifera blooms is achieved. Experiments on E.prolifera detection from multi-temporal SAR images are conducted for evaluation and results demonstrate our OORFC method's higher accuracy compared with the popular supervised classification methods (minimum distance and maximum likelihood). The capability of OORFC for E. prolifera bloom detection even during cloudy summer from SAR images can potentially facilitate the hazard management on the coastal environment.