Document Type: Research Paper
Oceanographic images obtained from environmental satellites by a wide range of sensors allow characterizing natural phenomena through different physical measurements. For instance Sea Surface Temperature (SST) images, altimetry data and ocean color data can be used for characterizing currents and vortex structures in the ocean. The purpose of this thesis is to derive a relatively complete framework for processing large dynamic oceanographic image sequences in order to detect global displacements such as oceanographic streams and to localize particular structures like motion current and vortices and fronts. These characterizations will help in initializing particular processes in a global monitoring system. Using area-based algorithms, two least squares methods have been used to solve the apparent motion which involves Least Squares Matching (LSM) and Hierarchical least squares Lucas and Kanade (HLK). SST images of Caspian Sea taken by MODIS sensor on board Terra satellite have been used in this study. Three daily SST images with 24 hours time interval are used as input data. The LSM technique, as a flexible technique for most data matching problems, offers an optimum spatial solution for the motion estimation. The algorithm allows for simultaneous local (i.e. template) radiometric correction and local geometrical image orientation estimation. Actually, the correspondence between two image templates is modeled both geometrically and radiometrically. In order to implement weighted least squares fit of local first-order optical flow constraints in each spatial neighborhood, the HLK method has been used. This method locates water current using coarse-to-fine strategy to track motion in Gaussian pyramids of SST images. This method allows the detection of large motion in coarse resolution layer and guides to correct result in finer layers. The method used in this study has presented more efficient and robust solution compared to the traditional motion estimation schemes to extract water currents.