Automatic marine targets detection using features based on Local Gabor Binary Pattern Histogram Sequence

In this paper, a new method for automatic ship detection using Local Gabor Binary Pattern Histogram Sequence (LGBPHS) is presented. In this approach, a ship image is modeled as a "histogram sequence" by concatenating the histograms of all the local regions of different local Gabor magnitude binary pattern maps. To detect ship targets, the input image is divided to overlapping blocks and based on the extracted features the ship area is detected. The extracted features are used to train an artificial neural network and SVM classifier for ship detection. The proposed algorithm is tested with different images containing single ship and without it and the results compared with those of using Haar-like features and cascaded classifier. The experimental results showed the proposed method is efficient. The method not only has proper result but also is robust against different imaging conditions.