## Hierarchical Clustering of Self-Organizing Maps for Cloud Classification (2000)

Venue: | Neurocomputing |

Citations: | 9 - 0 self |

### BibTeX

@ARTICLE{Ambroise00hierarchicalclustering,

author = {Christophe Ambroise and Fouad Badran and Sylvie Thiria},

title = {Hierarchical Clustering of Self-Organizing Maps for Cloud Classification},

journal = {Neurocomputing},

year = {2000},

volume = {30},

pages = {47--52}

}

### OpenURL

### Abstract

This paper presents a new method for segmenting multispectral satellite images. The proposed method is unsupervised and consists of two steps. During the rst step the pixels of a learning set are summarized by a set of codebook vectors using a Probabilistic Self-Organizing Map (PSOM, [9]) In a second step the codebook vectors of the map are clustered using Agglomerative Hierarchical Clustering (AHC, [7]). Each pixel takes the label of its nearest codebook vector. A practical application to Meteosat images illustrates the relevance of our approach.

### Citations

68 |
Pattern Recognition and Neural Networks (Cambridge
- Ripley
- 1996
(Show Context)
Citation Context ...zed by a set of codebook vectors using a Probabilistic Self-Organizing Map (PSOM, [9]) In a second step the codebook vectors of the map are clustered using Agglomerative Hierarchical Clustering (AHC, =-=[7]-=-). Each pixel takes the label of its nearest codebook vector. A practical application to Meteosat images illustrates the relevance of our approach. Key words: Kohonen maps, image segmentation, hierarc... |

33 |
Derivation of a class of training algorithms
- Luttrell
- 1990
(Show Context)
Citation Context ...r quantization method, topological maps introduces a additionnal relation between the codebook vectors. This \topological" relationship is a constraint which has been proved to produce robust results =-=[5]-=-. The most well-known topological algorithm is the Kohonen Self-Organizing Maps Algorithm. It assumes implicitly that the distribution to approximate is a Gaussian mixture where each component is de n... |

24 |
Cloud cover from high resolution scanner data: detecting and allowing for partially ®lled ®elds of view
- Coakley, Bretherton
- 1982
(Show Context)
Citation Context ...d thin cirrus above lower thick clouds. To resolve the ambiguities, the importance of information on the spatial variability of radiance eld such as the simple local IR variance, has been stressed by =-=[3]-=-. In this experiment we have considered four feature for describing each pixel: IR radiance, VIS radiance, IR local variance and VIS local variance (computed over a5 5 sliding window). 3.1 Results 4 C... |

15 |
Interpreting the Kohonen self-organizing feature map using contiguity-constrained clustering
- Murtagh
- 1995
(Show Context)
Citation Context ...ing probabilistic topological algorithms in a rst stage allow to summarize the initial data set into a smaller set of codebook vectors which can be clustered using a hierarchical clustering algorithm =-=[6]-=-. The topological algorithm tipically provides a partition of thousand clusters, which is used as starting point for hierachical clustering. Let us detail the two steps of this procedure. 2.1 Probabil... |

6 |
Statistiques et méthodes neuronales, Dunod
- Thiria, Lechevallier, et al.
- 1997
(Show Context)
Citation Context ...roposed method is unsupervised and consists of two steps. During the rst step the pixels of a learning set are summarized by a set of codebook vectors using a Probabilistic Self-Organizing Map (PSOM, =-=[9]-=-) In a second step the codebook vectors of the map are clustered using Agglomerative Hierarchical Clustering (AHC, [7]). Each pixel takes the label of its nearest codebook vector. A practical applicat... |

4 |
Cloud detection and analysis: a review of recent progress
- Goodman, Henderson-Sellers
- 1988
(Show Context)
Citation Context ...tlantic sea between the tropics. Many methodologies have already been developed for extracting cloud information from VIS-IR imagery of geostationary satellites, for an historical summary see [8], and=-=[4]-=-. In many cases, for an e cient separation in cloud type and an exact detection of the clear sky zones, the use of VIS/IR spectral signature is not su cient. From a top of the atmosphere observation, ... |

4 |
Global, seasonal surface variations from satellite radiance measurements
- Rossow, Brest, et al.
- 1989
(Show Context)
Citation Context ...er the Atlantic sea between the tropics. Many methodologies have already been developed for extracting cloud information from VIS-IR imagery of geostationary satellites, for an historical summary see =-=[8]-=-, and[4]. In many cases, for an e cient separation in cloud type and an exact detection of the clear sky zones, the use of VIS/IR spectral signature is not su cient. From a top of the atmosphere obser... |

2 |
Constrained clustering and kohonen self-organizing maps
- Ambroise, Govaert
- 1996
(Show Context)
Citation Context ...I. All the components are supposed to have the same covariance matrix. Generalisation of this algorithm have recently arised in the framework of this probabilistic interpretation of Kohonen algorithm =-=[2, 9]-=-. For example the authors [9] have considered that the distribution of the vectors z 2 IR d can be expressed as p(z) = X p(c)pc(z)� where p c(z) represent a Gaussian Mixture related to the codebook ve... |

1 |
Approche probabiliste en classi cation automatique et contraintes de voisinage
- Ambroise
- 1996
(Show Context)
Citation Context ...it can lead to biased results: some classes may benot represented or under sampled. In this paper a fully unsupervised approach is proposed. The method is based on Probabilistics Self-organizing Maps =-=[9, 1]-=- and works in two steps. The rst steps approximates the distribution of the pixels to be classi ed using a topological map. Hierachical clustering is then used for labeling the neurons of map. Preprin... |