Document for Meteorology Data Written by: Jia Li, The Pennsylvania State University, June 2010 ========================================================================= = Data creation: The data file DBZ_DATA_CONDITIONED.mat was created by Dr. Fuqing Zhang's Lab in the Department of Meteorology at Penn State University. It contains 41 cloud images generated for weather forecasting. ========================================================================= = 1. The data file is in matlab format. To load the data, use the following commands in Matlab: load DBZ_DATA_CONDITIONED; After loading the data, 41 arrays named DBZ_FT18_EF11 to DBZ_FT18_EF51 will be created. The value of each array element indicates the intensity of cloud at the corresponding pixel location. To view the array as an image by "xv", write out the data into an image file in pgm format. For example in Matlab: dat=DBZ_FT18_EF11; [row,col]=size(dat); f1=fopen('image.pgm','w'); fprintf(f1,'P5 %d %d 255\n',col,row); fwrite(f1, dat', 'uchar'); fclose(f1); Then use "xv image.pgm" to view the image under OS command. -------------------------------------------------------------------------2. These cloud images are used to test clustering algorithms that segment an image into clusters of clouds. The shapes of the cloud clusters which tend to be perceived by human vision are highly non-elliptical. This poses difficulty to many widely used clustering algorithms such as k-means or mixture-model-based clustering which implicitly assume Gaussian-type clusters.