Research On Optical Remote Sensing Imaging Characteristics Modeling And Detection Method Of Ship Targets In Complex Sea And Air Environment

Marine ship detection is of great significance to the military and civilian fields such as maritime traffic supervision,maritime search and rescue,marine defense in key ports,and mastering the allocation of enemy maritime resources.The continuous monitoring and detection of ship targets based on high-resolution optical remote sensing satellite images Has become a research hotspot in the field of optical remote sensing information acquisition and processing.However,it is interfered by the complex sea and air environment such as atmospheric transmission,cloud and fog,different sea conditions in the optical remote sensing image acquisition link,and the high-resolution optical remote sensing image may contain various types of ship targets,large aspect ratio,random directions and ports Due to factors such as dense nearby targets,existing methods are difficult to meet the application requirements of low false alarm and high probability ship target detection.In this paper,to solve this problem,we carry out research on modeling and detection methods of optical remote sensing imaging characteristics of ships in complex sea and air environment,and carry out corresponding experimental verification.The main research work is as follows:(1)Modeling and characteristic analysis of optical remote sensing imaging link in complex sea and air environment.First,by modeling and characterizing the structure and radiation characteristics of the ship's target body,ship's target wake,complex sea background and cloud background,the modeling and simulation of the characteristics of typical target scenes are realized;then,based on the optical remote sensing imaging mechanism,the imaging is analyzed Influencing factors of optical remote sensing imaging characteristics such as atmospheric transmission,satellite platform,and optical load in the link,and establishing a characterization model of optical remote sensing imaging characteristics of ship targets in complex sea and air environments.Finally,the parameters of the orbiting satellite system are used as input conditions to generate simulation The remote sensing image data of ship targets in different application scenarios is presented.(2)Ship target detection method and performance analysis based on artificial features.Firstly,combining the characteristics of optical remote sensing imaging of ship targets in complex sea and air environment,the characteristics of gray,color,texture and shape of ship targets in optical remote sensing images are analyzed.The results show that the grayscale and shape features have a certain ability to characterize the difference between the target and the background in the weak clutter interference image.Then,the ship detection method of artificial features was used to test the optical remote sensing image of complex sea and air environment.Experimental results show that low-dimensional visual features such as gray scale and shape cannot meet the robust description of targets in complex environments.In order to achieve low false alarm and high probability ship target detection,deeper extraction and higher levels of low-dimensional visual features are required.Characterization of dimensions.(3)Research on ship target detection method based on deep learning features.Use the high-dimensional data modeling capabilities of deep neural networks to characterize targets in higher dimensions,and design cross-scale connected feature pyramid networks to achieve multi-type,multi-scale,densely arranged ship targets in complex sea and air environments Feature extraction,for the random problem of ship target angle,the positioning method of rotating bounding box is used to achieve high precision and high robust calibration of target position and contour.Then,for the feature mismatch problem,the feature map is reconstructed by the feature refinement module to obtain the ship target detection network in complex sea and air environment,and the ship target detection method based on deep learning features is proposed,which realizes the complex sea and air environment more effectively High probability detection of ship targets in the background.(4)Experimental verification and analysis.Using HRSC2016,DOTA and other public data and simulation data to build a test data set to verify the correctness and effectiveness of the target detection method of the complex sea and air environment ship in this paper,combined with the performance evaluation index of the target detection algorithm,the analysis of this method in different degrees of background Target detection performance under wave interference.Experimental results show that the method in this paper can effectively realize the high probability detection of ship targets of optical remote sensing images in complex sea and air environments such as different waves,clouds and ports.