A Novel Nearest Feature Learning Classifier for Ship Target Detection in Optical Remote Sensing Images

Satellite remote sensing data is becoming more and more abundant, In order to realize automatic detection of ships on the sea surface, this paper presents an adaptive intelligent ship detection method, a novel nearest feature learning classifier (NFLC), which combines the scale invariant feature transform (SIFT) feature extraction with nearest feature learning classification. Due to the wide variety of detection ships, the NFLC can obtain a better experimental result than conventional detection methods. The detection accuracy is enhanced by the feature training in large databases and the performance of the system can be continuously improved through the target learning. In addition, compared to convolutional neural network algorithm, it can save the computation time by using the nearest feature matching. The result shows that almost all of the offshore ships can be detected, and the total detection rate is 89.3% with 1000 experimental optical remote sensing images from Google Earth data.