Conjunctive fusion by operator ``min'' in possibility theory corresponds to the fusion by the Bayes' rule in the field of probability theory. These fusions are adapted to the aggregation of sources. It is the case here, since the spectral bands can be considered as images of a scene taken at the same time and almost in the same position. Each pixel x of a spectral band I_i corresponds thus to the same pixel x of the spectral band I_j.

So, all the spectral bands agree, at the beginning, to say more or less strongly that pixel x belongs to class c. Information to be fused are redundant.

However, the sensors used for the shots are not perfect, because the images they provide contain additional noise. Moreover, one pixel is an amalgam of the various objects present at the surface of the ground within the limit of the space resolution of the sensor used.

For these images, the sensors MSS of LANDSAT have a space resolution of 57 meters (line) x 79 meters (column). A pixel thus contains information made up of the various spectral signatures of the objects within the limits of the pixel. But the various sensors are not sensitive in the same way to the different spectral signatures. For example sensor MSS7 (near infrared) is very sensitive to the presence of water, and recognizes class 7 (irrigated cultures) in the best way.

These various difficulties make more difficult the fact that all the spectral bands affirm that a pixel x belongs to class c.

• Principle
• Conjunctive Fusion of the Four Spectral Bands MSS4 to MSS7
• Conjunctive Fusion of the Four Spectral Bands MSS4 to MSS7 and of Out-Image Data
• Conclusion on Conjunctive Fusion

                  
  
 IRIT-UPS