Topographical matching of multisensor information

Discusses the n-dimensional vector representation of the properties of multitemporal or multispectral remotely sensed objects. The known classification procedure permits object identification as i-th class in n-dimensional space. For many cases, increasing n leads to higher informativity and thus to higher probability of correct identification. The number of information channels being given, higher informativity can be reached using multitemporal sensing. This technique is especially productive for vegetation study, because each kind of vegetation has its own temporal gradient of properties evolution, even on comparatively short time intervals. For multichannel sensing both for simultaneous multispectral and multitemporal ones, spatial correspondence is the problem of principle. The authors describe an image classification scheme which consists of two stage pixel to pixel identification. The first stage provides topographical coincidence of the same point under analysis for all channels and the second stage results in characterisation of the pixel under study