| As countries pay more and more attention to the protection and developm ent of marine resources,the detection methods of ship targets have great resear ch value in many fields.There are few studies and,most of them are limited to the background of sea and air.Compared with traditional detection methods,detection method based on infrared has great application value in ocean monito ring,especially in weak target signals.At present,there are relatively few studi es on space-based infrared ship detection,and the research on target characteris tics is not deep enough.Due to the lack of space-based infrared high-resolutio n ship data,infrared ship target detection can only be studied at low and medi um resolutions.This paper focus on ship target under the conditions of the low resolution of space-based infrared image,take the quantitative detection as the research object.The quantitative detection is about geometric features and radiation featur es.Target geometric quantitative is embodied in the multi-scale and target geo metry characteristics under different sampling system is analyzed,according to the different scales of target to evaluate system indexes such as resolution and sampling system.The quantification of radiation characteristics is reflected in a ccurately calculate the radiation characteristics of different infrared spectral ban ds of the target,hoping to provide certain support and reference for the whole process of satellite target detection,identification,inversion and performance op timization.The main research are as follows:(1)For the thick clouds and coasts,the areas which have higher energy i n the medium and short infrared wave remote sensing images and more influe nce on the target detection are detected and separated.The division of the land is accomplished by means of superpixels that are characterized by a prominen t boundary on the image and a sensitive feature of the ground.The support vector machine(SVM)is used to classify the grayscale multi-dimensional features of the image,and the thick cloud is more accurately identified.(2)Aiming at solving the problem that the complex background weak sign al of infrared ship small target image is difficult to be accurately detected,a CFAR(Constant False-Alarm Rate)-DCRF(Dense)based on constant false alar m rate and fully connected condition random field is proposed.The algorithm i s based on the characteristics of small target and false alarm signal changes bu t different structural characteristics.It uses the advantages of multi-dimensional context(space,radiation)expression of conditional random field(CRF)to reali ze false alarm feature suppression and introduces CFAR to model to improve t he detection ability of weak signal targets,and achieve accurate detection and segmentation of small ship targets.(3)The geometric characteristics such as the length,width,shape and dire ction of the target are the basis for the accurate identification of the target afte r detection.The geometric features are different in remote sensing images with different resolutions and different sampling systems,which affects the recognitio n accuracy.Based on the camera model,this paper establishes the imaging link simulation under multi-resolution conditions,analyzes the ship target structure and direction characteristics under different scales and different sampling system s,and evaluates the system indexes such as resolution and sampling system.T he analysis results show that for ships with a length of 100 meters,the shape and direction characteristics can be stable when the spatial resolution is higher than 12 meters.Compared with conventional sampling,oversampling can impro ve the spatial resolution of the image,but the target edge.The diffusion is mo re obvious,which is not conducive to the extraction of target shape features.For a ship target with a symmetrical structure,the two sampling systems h ave the same ability to capture and judge the direction of the moving target.T his conclusion provides a certain reference for the optimization of system indic ators.(4)In terms of radiation characteristics,this paper analyzes the infrared target radiation characteristics of different bands,and proposes a separation model for the problem of reflection and emission coupling of medium wave targets.The equations are calculated based on the physical radiation model.The relative error of the target brightness temperature inversion is 1%,which improves the accuracy of the target radiation characteristics... |
PDF Full Text Request