Ship recognition and classification using silhouettes extracted from optical images

In this paper, extraction of ship signatures from silhouette images of three-dimensional ship models and ship recognition from optical images are investigated. First of all, from the silhouette images of 3-dimensional ship models, with the help of feature vectors, ship signatures are created. Using three-dimensional ship models gets rid of the difficulty of obtaining real videos for the database and makes it possible to obtain information about ships from every angle. Then, created ship signatures are collected in a synthetic database. In the next stage, using segmentation and Artificial Neural Networks, ship recognition and classification are performed. In this paper, how all of the stages are done is explained and obtained numerical results are provided to illustrate used theoretical solutions.