Super-Pixel Fisher Vector-based Green Algae Detection in Multispectral Remote Sensing Images

In this paper, we propose a novel green algae detection method that incorporates both statistical and spatial characteristics of multispectral remote sensing images. First, superpixels are used as the processing elements to exploit the spatial characteristics of multispectral remote sensing images. Second, we model the statistical distribution of superpixel intensities using the Gaussian mixture model and calculate Fisher vectors (FVs) for each superpixel to extract the multi-order features therein. Finally, the random forest classifier is applied to classify FVs of superpixels in multispectral remote sensing images into green algae and non-green algae areas. Experimental results demonstrate the superiority of our proposed method in comparison with existing methods using measured data.