英文摘要:Man-made buildings detection is important in land use supervision and land control applications. Generally, polarimetric synthetic aperture radar (PolSAR) data are processed to detect buildings well. But for some buildings which are not aligned with the radar track, these buildings are usually incorrectly recognized as forest, because the oriented buildings produce additional cross-polarization. Polarimetric interferometric SAR (PolINSAR) acquires two measurements with a spatial baseline or a temporal baseline. For the PolINSAR with a temporal baseline i.e., the repeat pass PolInSAR, the two polarimetric measurements are sensitive to targets' temporal variations during the time. The buildings, regardless of the orientations, have high coherence, while some natural targets have low coherence. A novel parameter is proposed here, which represents the mean PolINSAR coherence and can be utilized to distinguish between buildings and some natural targets. The parameter is based on the coherence optimization theory of Cloude and Papathanassiou, and is the mean of the three optimal coherences with three pseudo-probabilities. Based on this new parameter and the SPAN, a method to detect buildings is further proposed. The excellent performance of the proposed method on buildings extraction is demonstrated by processing German Aerospace Center (DLR) L-band E-SAR repeat pass PolINSAR data of Oberpfaffenhofen area. �2016 IEEE.