Study On The Ship Detection In SAR Image Based On CFAR And Deep Learning

With the development of the Marine exploration, Marine transportation, and Marine resource exploitation, more and more frequent marine ships activities is getting.There is very important practical significance and strategic significance for the country's social economic development and the maintenance of Maritime rights and interests to detect the position of the ship at sea accurately and quickly.SAR is an important method to implement ships detection, with the day and all-weather characteristics. CFAR algorithm is one of the most widely used algorithms in the field of ship detection, however, with the rapid development of satellite remote sensing technology, remote sensing data gradually show the characteristics of "big data", and ship detection also need to improve its accuracy and speed by combining large data analysis .The deep learning has made a breakthrough in the use of large data for computer pattern recognition. Among them, the convolutional neural network (Convolutional Neural Network, CNN) is the best architecture in static image recognition effect. In this paper, the following three ship target detection schemes in maritime SAR images are studied and realized respectively, on the basis of CFAR algorithms and CNN model.At first, the multistage CFAR detection algorithm based on multithreading technology is studied and realized to achieve rapid ship detection. Through contrast research and analysis of the used CFAR detections in the statistical model, the most suitable model through them is screened out for the distribution of sea clutter. The optimization scheme of CFAR algorithms is provided based on the classical CFAR detection idea, and applied in ship detection.Secondly, the ship detection algorithm based on CNN is studied and realized in view of the ship detection accuracy. Based on the analysis of CNN model structure, principle and the whole training process, in combination with the classic case of CNN model in image classification,the ship detection scheme based on deep CNN is studied and put forward,whose experimental study in ship detection is carried out.Finally, the ship detection algorithm research in the combination of CFAR and CNN is carried out in view of the ship detection speed and accuracy. Through analyzing the detection results of the front two ship detection scheme, it is shown that the detection speed of multistage CFAR algrithm based on multi-thread technique is improved, but the detection accuracy not high. The detection accuracy of the CNN algorithm is improved, but the detection time long. The ship detection scheme in the combination of CFAR and CNN is designed and realized,by using for reference the global detection in multistage CFAR algorithm based on multithreading technology and the optimization model in CNN algorithm.