A novel vessel detection and classification algorithm using a deep learning neural network model with morphological processing (M-DLNN)

In recent times, optical satellite images are gaining widespread significance due to its feasibility and compatibility with laser images to improve the contrast of the obtained image. Optical satellite images find numerous applications in object detection and tracking out of which ship detection and tracking is a quite significant field to be researched. A novel ship detection and tracking algorithm has been proposed and implemented in this research article and tested over a wide range of climatic conditions to quantify the efficiency of the proposed work. The proposed algorithm defines morphological traits of the object under study which is the ship in this case as its attributes to segment and detect the object. An input video sequence obtained from the optical imaging system is used as input in the proposed work and efficiency justified in terms of accuracy, precision and classification rates. A morphological deep learning neural network has been proposed in this research article which automates the function of both detection based on morphological attributes of the vessel under study and classifying the detected vessels into their constituent classes. The obtained results have been compared with recent and existing vessel detection algorithms like singular-value decomposition, salient mapping techniques and local binary pattern-based techniques.