Fast Detection And Recognition Of Maritime Targets In Visible Remote Sensing Images

In recent years,with the rapid development of remote sensing detection technology,a variety of sample data have been provided for marine target detection and identification,and remote sensing images have been increasingly applied in military and civil fields.According to the actual needs,in order to monitor and control the key sea harbor area of ship flow and distribution of surface ships,the identification of maritime targets is divided into harbor area detection and ship target identification.Focusing on the automatic detection and recognition of harbor and ship targets in visible remote sensing images,edge processing,feature extraction and target detection are studied in the harbor detection part,saliency detection and target classification are studied in the ship identification part.On this basis,the automatic detection and identification methods of harbor and marine ship targets with high detection accuracy and low false alarm rate are designed respectively,improve the sea target detection and recognition technology of remote sensing image efficiency and real-time performance.The main research contents of this paper are summarized as follows:1.The basic principle of harbor and ship target detection and recognition under visible light remote sensing image is introduced,and the main characteristics such as shape,color and texture of harbor and ship target are analyzed in detail,providing a theoretical basis for the research of marine target detection and recognition algorithm.This paper summarizes the existing harbor and ship target detection and recognition algorithms respectively,analyzes the advantages and disadvantages of all kinds of algorithms,and introduces the basic theory of the key technology of the detection and recognition algorithm in detail.2.A remote sensing image processing technology based on edge preserving algorithm is proposed,which can effectively reduce background interference of remote sensing images.In the remote sensing images containing harbors,due to its proximity to the coastline,there will be many man-made structures and natural features near the coast,which will interfere with the feature extraction of harbor targets.Through experimental evaluation,the edge preserving algorithm proposed in this paper can effectively eliminate these complex and useless background information,and can effectively retain key edge information,improve the detection accuracy of feature extraction and target recognition.3.A SIFT feature extraction algorithm based on edge classification is proposed,which can effectively reduce the extraction of non-edge features.In view of the fact that harbor target features belong to obvious edge features,the algorithm designed in this paper can focus the extracted features more on key feature points,reducing the mismatching rate of harbor target identification;4.A saliency detection model based on hypercomplex Fourier transform is proposed,which can quickly search for information related to the current task requirements.This paper discusses the research value of visual saliency model and the status quo of the application in the field of target detection,at the same time,the construction of visual saliency models such as saliency target detection and its application in visible remote sensing image ship target detection are studied in depth.An improved hypercomplex Fourier transform saliency model is designed to locate the target area of ships on the sea,which can effectively suppress the background interference such as clouds and sea clutter.5.A Res Net model based on transfer learning is proposed to identify the ship targets extracted after saliency detection,which solves the problems of gradient disappearance and gradient degradation when the CNN network layer is too deep.The scheme based on transfer learning can train the CNN model with a small amount of ship sample data and realize the classification of ship targets with higher accuracy.To sum up,in this paper,the problems and challenges in the field of automatic detection and recognition of marine targets in visible remote sensing images are analyzed in detail,according to the key technical problems such as harbor target feature detection,area of interest extraction,target detection,as well as the saliency model,target detection and identification of marine targets are studied and some results are obtained.The relevant achievements in this paper can provide theoretical basis and algorithm support for automatic detection and recognition of harbor and ship targets in visible light remote sensing image in aerospace field,and have reference significance.