Research On Airplane Classification And Recognition From SAR Image

Aircraft is one of the important military targets,fast and effective classification and recognition of aircraft targets in SAR images is an important task of battlefield intelligence reconnaissance.Like other maneuvering target ATR SAR systems,a typical SAR image aircraft target recognition system includes three steps,including detection,identification and classification.And target classification is the ultimate goal of SAR image target recognition,which is a very important part of ATR SAR.In order to carry out the classification of aircraft targets,this paper adopted a method of hierarchical processing,the classification process is divided into SAR image preprocessing,feature extraction,target detection of airport aircraft,aircraft target geometric feature extraction,classification and recognition of five aircraft each step.First,remove the speckle in SAR image.Then,quickly extract the airport area and have a detection of aircraft targets in the airport area.Next,based on the results obtained from the aircraft target detection,the geometric features of the aircraft target are extracted.Next,based on the results obtained from the aircraft target detection,the geometric features of the aircraft target are extracted.Finally,the geometric characteristic value of aircraft target is used to classify and identify it.The main work of this paper includes the following aspects:(1)In this paper,the extraction method of the airport runway was studied,and the results of airport target extraction were obtained by using the method of growing points.Finally,the airport area of the aircraft target was detected.CFAR detection with a high false alarm rate is used for rough separation of potential targets at the airport,and then use the pixel clustering and area selection,remove the false target detection results,get the final segmentation result of aircraft.(2)This paper presents a method of aircraft target feature extraction based on the variable geometry model of SAR image.In the SAR image,the contour of aircraft target is complex,and the regional connectivity of the aircraft target is poor.This method only needs to determine the range of parameters of aircraft geometric model according to the aircraft's prior knowledge.Then,the optimal geometric model of aircraft target is automatically fitted by genetic algorithm,which can extract the geometric features of SAR image,which provides a useful way for SAR image target.(3)In this paper,based on the extraction of the geometric characteristics of aircraft targets in SAR images,the method of D-S evidence fusion is used to classify the aircraft targets.In the process of classification,the characteristic value of the aircraft target is used as the evidence of the classification of the target.The classification results of the target are obtained by each evidence,and then the final results of the classification of the SAR images are obtained by using some decision rules.