Methods Of Infrared Ship Detection And Target Recognition

Ocean that covers 70% area on earth is one of the most significant military bases which cannot be ignored for countries, especially as the world military power rapidly developed,various countries pay greater attention to naval warfare. With imponderable advantages, the detection and recognition technology of infrared target occupies an important place in the field of military. It is continually researched and innovated. Considering ships are the main infrared target in the sea area, the detection and recognition of infrared ships is particularly significant, and has great military applications and research value.This thesis aims to deeply research the detection and identification of infrared ship images. Versatile detection method is obtained for the analysis of characteristics of complex infrared ship images. Identification of the target ship also carried out a systematic study. The discrimination, which has been studied in the past, of ship type is refined to identify the different categories of ship models. A series of efficient detection and recognition methods of infrared ships are worked out, and the effectiveness of the proposed methods is verified by simulation. This thesis studies on the following aspects:Firstly improvement to traditional P-M model enables the processed images to retain edge features and details of the full retention of infrared ship target under the premise of a good noise suppression, and unsupervised smooth of P-M model can be achieved by threshold attenuation;Secondly in order to solve the issues of high sensitive for initial outline, GVF-Snake model contour extraction method based on PCNN is obtained, with the optimization of dealing with the disordered distribution in gradient-diffusion field of GVF-Snake model, after the comparison and research among outline initial methods based on OTSU algorithm,SUSAN algorithm and PCNN algorithm, to achieve regularization of gradient vector field;Thirdly a new specific geometric feature is adopted creatively to display the characteristics of different types of ships. The final features that combine the new feature and traditional geometric feature with invariants feature characterize the ship targets, intensifying the distinction among a variety of integration features;At last Extreme Learning Machine is applied to identify and classify the infrared ships characteristics with its improved model for better adaptability of small sample training and identification.Combining all the above-mentioned method, all the involved algorithm are demonstrated for the detection and identification process of infrared ships with MATLAB platform GUI.