Field boundary detection using multi-temporal SAR

A new methodology for automatic field boundary detection based on multi-temporal SAR data is presented. A method was developed with the consideration of operational usability while applying on ERS-2-SAR-SLC data of a region north of Munich. The aim of this work is to develop a technique that combines the advantages of high precision GIS with weather independence and fast availability of remotely sensed radar data. High precision coreferencing of SAR-SLC data as used for interferometric techniques enables the application of eigenvector based methods, such as principal component transformation and singular value decomposition. These have been applied in an adaptive way for better results on scenes including different geographical regions. The application of a nonlinear Gaussian filter on transformed data is applied to reduce speckle and increase perceptibility of field boundaries. This enabled fast segment based classification taking regional aspects out of GIS into account