Ship Detection And Recognition Based On Optical Remote-Sensing Images

Ship detection and recognition has been a popular research topic thanks to the importance of ship targets.With the development of the optical satellite remote sensing technology,ship detection and recognition based on optical remote-sensing images has broad application prospects.Driven by it,this paper did research in three aspects,including ship detection,ship recognition and software system development.The main works of this paper are as follows:1)Water expansion.The interfere of land background in the optical remote-sensing image could affect the results of ship detection.Grayscale threshold segmentation is adopted for the sea-land segmenting of the offshore ships situation,and templates matching method utilizing harbor prior knowledge is adopted for the sea-land segmentation of the inshore ships situation.Then the mean values of water pixels are used to replace the values of land pixels for water expansion.2)Ship detection.Based on the investigation of state-of-art detection a lgorithms,visual saliency algorithm is introduced in consideration of the complicated background of ships.Two ship detection algorithms are proposed in this paper,one is based on itti and RX,and another is based on MSSS+wCtr and RX.The proposed algorithms can improve the precision of detection and reduce the false alarm rate.3)Ship recognition.This paper adopts the coarse-to-fine strategy to solve the ship recognition problem.First,a pre-trained SVM classifier is used to eliminate the false targets.Then,a pre-trained Adaboost classifier is applied for target classification.4)Ship detection and recognition system.Based on above methods,a system for ship detection and recognition is designed.Detailed design of the system in cluding water expand model,ship detection model and ship recognition model are given in this paper.