Change detection in SAR-images from the TerraSAR-X satellite with main focus towards harbors and container terminals

The detection of changes or no change in a set of images of the same scene recorded at different times is of importance in many applications. The time scale could be from seconds to days or months. Man-made constructions often contain regions with high scattering from radar waves that are readily detected. The use of Non-Coherent Change Detection (NCCD) from intensity images or additionally using the phase information in the complex SAR images in a Coherent Change Detection (CCD) are powerful techniques that can detect changes and even repositioning of objects in almost the same position. This report presents results using both techniques based on satellite SAR images. This exemplifies what can also be obtained by an airborne platform with proper control of radar and navigation. Based on two high resolution satellite images recorded of the Oslo harbor region analysis has been performed to extract information of changes. The two images were recorded with 11 days interval and were co-registered to a fraction of a pixel. The high quality of these images represent an example of what would be the best performance obtainable given that precise control are available in an airborne SAR platform. Analyses have been performed for NCCD and CCD in the two images. Different objects have been studied with a detailed analysis of container terminals and what information could be extracted from such objects. Simulations of SAR images using modeled containers in different stacking configurations are also used to help understand the backscattered response from containers. Based on this knowledge, changes to container stacking configurations were extracted from the images. A coloring scheme was applied to easier extract changes in the CCD analysis. The two magnitude images and the coherence measure from the CCD calculations were used in the R, G and B channel of an image, respectively. Objects only present on day 1 or 2 respectively appear as red or green colored objects. An object with high backscattering that are unchanged from day 1 to day 2 and that result in high coherence will appear as white. A series of objects are presented that show how these techniques can provide powerful tools for change detection.