Automatic Ship Detection Using CFAR Algorithm for Quad-Pol UAV-SAR Imagery

Remote Sensing data, either airborne or satellites, are very much useful for incorporating the Geographical Information System (GIS) technology. SAR sensors are good as compared to optical sensors for monitoring maritime activity due to their capability of penetrating clouds and can work without depending upon any weather condition. SAR sensors can work day and night while optical sensors need a source to illuminate the surface hence can only work in the daytime. Many studies have been done on UAV SAR sensors for different applications like oil spills, ship detection, etc. Moreover, the polarimetric technique helps in understanding the feature much more in detail by using phase information like orientation and shape of the object using scattering behavior. In this paper, the main focus of the study is the Automatic ship detection using the Adaptive Threshold Algorithm popularly known as Constant False Alarm Rate (CFAR) for polarimetric UAV SAR data. Coherency Matrix (\({T}_{3}\)) is computed from quad-pol covariance SAR data \({C}_{3}\) and CFAR algorithm is applied to each element of the coherency matrix to detect ships. The sea surface follows the surface scattering and this can be highly helpful to distinguish the ships from the sea background. Moreover, due to the homogeneous background of imagery, the CFAR algorithm works more precisely as it can compute the adaptive threshold for each pixel using the background area by assuming it to the Gaussian in nature. Moreover, the Global Self-consistent, Hierarchical, High-resolution Geography Database (GSHHG) vector coastline layer and Digital Elevation Model (DEM) are used for masking out the land area to enhance the area of interest. In this study, \({T}_{22}\) element of the scattering matrix shows better results in the detection of the ships and in determining the shape of the ships. Finally, the efficiency of the algorithm is measured using the Receiver Operating Characteristics (ROC) curve.