The Research And Implementation Of SAR Mission Adaptability Evaluation Based On TWSVM

With the increasing application of synthetic aperture radar(SAR)in the field of information reconnaissance,reducing the use cost and studying its mission adaptability are the problems to be addressed.Traditional adaptive assessment methods are mainly qualitative analysis,which require high labor cost and have no further analysis of the assessment results.The focus of this paper is to complete the adaptive evaluation of SAR image.Aiming at the problem of imbalance in the number of samples in actual scenes,this paper simulates SAR images under different flight scenarios through the airborne SAR mission adaptability evaluation system,and uses SDAE and WGAN-GP to build models to improve the impact of biased data on the evaluation results.For the data-balanced sample set,this paper analyzes the airborne SAR image data set,integrates the traditional quality evaluation index,texture evaluation index and edge continuity evaluation index to form the task adaptability evaluation feature vector.This paper uses the twin support vector machine algorithm(TWSVM)to carry out airborne SAR mission adaptability assessment.The main work of this paper is as follows:(1)Existing computing resources can not meet the requirement of directly generating SAR images.In this paper,the SDAE network model is used to extract image sample features.Comparing the method of using PCA to extract features and the method of using SAE to extract features,the method of using SDAE to extract features has the highest accuracy,increasing by 3.5% and 3% respectively.(2)Aiming at the problem of the imbalance in the number of samples passing and failing to evaluate in real task scenarios,this paper designs a task adaptability evaluation method based on SDAE and WGAN-GP,and constructs a feature extraction layer,sample generation layer and adaptability,which is three-layer network model of the evaluation layer.With the increase of the imbalance ratio,compared with the method of using SMOTE for feature sample expansion,the decrease of the G-mean value of the method in this paper is smaller than the decrease of the G-mean value of the SMOTE method.When the imbalance ratio is 100:1,the G-mean value of the method in this paper is 3.9%higher than that of the SMOTE method.(3)Based on the traditional SAR image quality evaluation method,this paper designs a SAR task adaptability evaluation method based on TWSVM.Based on the existing SAR image quality evaluation index system,this paper adds texture evaluation indexes and edge continuity evaluation indexes to evaluate the adaptability of SAR tasks from many aspects.In order to analyze the adaptability evaluation results,this paper proposes a credibility evaluation method based on the TWSVM algorithm.Compared with the traditional evaluation method,the accuracy of the evaluation method in this paper has increased by 3.7%,and the calculation speed has increased twice.The credibility evaluation introduced has guiding value for subsequent processing.