A Multiscale Approach Based Automatic Ship Detection

Abstract

This paper presents a new method for ship detection. It uses SAR image obtained by passing pulses of microwave signal and recording the reflected signal. SAR image is used for ship detection because it functions irrespective of weather conditions even at day and night. Speckle noise is inbuilt in SAR images are formed by recording the radar echo signals of EM signal multiplicative with granular noise;this noise degrades the quality of SAR images. Preprocessing is done using Lee filter to remove the speckle noise. Quality measurement like SNR, PSNR, SSIM and ENL is done. Image profile is verified to find effective noise suppression. After pre-processing the SAR image is enhanced using DWT (discrete wavelet transform) this wavelet transform preserves high frequency information and reduces artifacts. Enhanced image consists of many information from which the relevant image has to be extracted this process is called as feature extraction, feature extraction is done using k-means Clustering algorithm were the given data is grouped into k number of clusters. Finally, the ship detection image is obtained and it is compared with the ground truth image, quantitatively the ship Detection image is evaluated using confusion matrix.