Automatic Interferogram Selection for SBAS-InSAR Based on Deep Convolutional Neural Networks

Abstract

The small baseline subset of spaceborne interferometric synthetic aperture radar (SBAS-InSAR) technology has become a classical method for monitoring slow deformations through time series analysis with an accuracy in the centimeter or even millimeter range. Thereby, the selection of high-quality interferograms calculated is one of the key operations for the method, since it mainly determines the credibility of the deformation information. Especially in the era of big data, the demand for an automatic and effective selection method of high-quality interferograms in SBAS-InSAR technology is growing. In this paper, a deep convolutional neural network (DCNN) for automatichigh-quality interferogram selection is proposed that provides more efficient image feature extraction capabilities and a better classification performance. Therefore, the ResNet50 (a kind of DCNN) is used to identify and delete interferograms that are severely contaminated. According to simulation experiments and calculated Sentinel-1A data of Shenzhen, China, the proposed approach can significantly separate interferograms affected by turbulences in the atmosphere and by the decorrelation phase. The remarkable performance of the DCNN method is validated by the analysis of the standard deviation of interferograms and the local deformation information compared with the traditional selection method. It is concluded that DCNN algorithms can automatically select high quality interferogram for the SBAS-InSAR method and thus have a significant impact on the precision of surface deformation monitoring.