Ship Detection from Satellite Images using Deep Learning

The features of synthetic aperture radar (SAR) has been widely used in maritime surveillance. While various object detection techniques have been proposed, currently, the techniques for detecting ships are inconsiderate to the small-scale vessels. Due to the complexity of images, traditional classification techniques often fail to classify complex images. Also, the intricacies of numerous ships, varying imaging settings, and limited tagged images have made ship categorization difficult in optical images. While the traditional classification approaches rely on feature extraction; however, they frequently fail to design well-performing features for complicated images,this paper presents a method that learns discriminative features and achieves strong classification accuracy using deep networks. A novel multi-scale method that uses an R-CNN network has been applied to extract the features of the images. We studied various aspects of ship photography and present the findings in the experiments. The R-CNN-based technique works well for ship categorization, capable of learning discriminative features and attaining strong classification accuracy using deep networks. The proposed work has been undertaken numerous investigations and produce two minor datasets of optical ship pictures for training and test data. The results of the experimentshave shown that the proposed method is effective for ship recognition than the state of the art approaches.