Extraction methods for sar imagery: A review for road features

SAR datasets provides an abundance of information as it has high penetration capacity. Varioustechniques for feature extraction has been developed rapidly in the recent times. Linear Features such as roads,canal, pipeline etc. plays a major role in city planning and development. The incessant development of roads inrecent times has raised the amount of labour and finance required for maintenance of roads. Information fromSAR imagery such as intensity, phase, backscattering, polarization, etc. is used to identify features dependingupon its incidence angle and dielectric constant. SAR has a requirement of pre-processing of data before extractionof any useful information. This paper summarizes multiple methods for Road Feature Extraction such as Markovrandom field (MRF), Bayesian Tracking Framework using particle filter method, Multiple Views of imagery withdifferent flight direction, Deeply Convolution Neural Network using Binary Segmentation and Regression,Various Transformation and Segmentation techniques etc. Various datasets required to process above extractiontechniques and its application has also been reviewed. Comparison of various models with its advantages anddisadvantages has also been performed. As, a single feature extraction technique will not give satisfactory resultso clubbing of different methods can be done based on its characteristic and need. This can be beneficiary to userswho are interested in road network and its analysis.