Study on Edge Detection of Multispectral Remote Sensing Image in Multidimensional Cloud-Space

An edge detection algorithm of multispectral remote sensing image based on multidimensional cloud-space was proposed. Firstly, we gain the central area of spatial object by region growing algorithm. By calculate the average gray level of central area as feature of cloud core, we propose multidimensional cloud-space mapping model. After mapping image into cloud-space by this model, the objects could be transformed into object-clouds with digital characteristics. By logical operation between intersecting clouds and the synthesis of multidimensional vectors, we could gain the digital characteristics of boundary-cloud to determine the reasonable scope of transition region. Deducing the membership of each feature point belongs to boundary-cloud by the digital characteristics of object-clouds, the fuzzy feature matrix was proposed. We could gain the optimal threshold by scanning this matrix based on maximal fuzzy entropy rules. We conduct various performance studies show that this algorithm is both efficient and effective.