Combining Color Fractal with LBP Information for Flood Segmentation in UAV-Based Images

The paper presents a method for patch classification to the end of flooded areas segmentation from aerial images. As patch descriptors color fractal dimension and color local binary patterns were proposed, so both color and texture information is combined. The remote images were taken by the aid of an Unmanned Aircraft System (MUROS) implemented by an authors' team. The algorithm of remote image segmentation has two phases: the learning and the segmentation phase. The class representative consists of a set of intervals created in the learning phase. The classification is made by a voting criterion which takes into consideration the weights calculating from both descriptors. The results were obtained on 100 images with high resolution from the orthophotoplan created with the images taken in a real mission. The accuracy of segmentation was better than in the separate approaches (single fractal or local binary patterns descriptors).