Segmentation for detecting buildings in infrared space images

摘要

The given work describes a new technique of image segmentation, in particular for building detection on radar or infrared Earth-observation images. The method is based on property of most man-made objects which consist in straight edges and mostly right angles. The developed 2D adaptive image filter assists to detect straight edges even if given image fragment has a low contrast and has been extremely noised, in addition, if an object edge has been distorted, for example, by interference in the SAR azimuth channel, the filter compensates for distortions which do not exceed the specified value. The next processing of line-segment list without image raster works faster and allows detecting a relatively small set of possible targets. This approach could be used as addition for neural networks as well as provide assistance in preparing of training data set.