ICA and GA feature extraction and selection for cloud classification
ICAPR'05 Proceedings of the Third international conference on Advances in Pattern Recognition - Volume Part I
| Hi-index | 0.00 |
In this work a back propagation neural network is used for thesegmentation of Meteosat images covering the Iberian Peninsula. Theimages are segmented in the classes land (L), sea (S), fog (F), lowclouds (C_L), middle clouds (C_M), high clouds (C_H) and cloudswith vertical growth (C_V). The classification is performed from aninitial set of several statistical features based on the gray levelco-occurrence matrix (GLCM) proposed by Welch [1]. This initial setof features is made up of 144 parameters and to reduce itsdimensionality three methods for feature selection have beenstudied and compared. The first one includes genetic algorithms(GA), the second is based on principal component analysis (PCA) andthe third uses independent component analysis (ICA). These methodsare conceptually very different. While GA interacts with the neuralnetwork in the selection process, PCA and ICA only depend on thevalues of the initial set of features.