Research On SVM-based SAR Image Optimization Recognition Algorithm

SAR(Synthetic Aperture Radar),that is,synthetic aperture radar,has unique advantages in disaster monitoring,environmental monitoring,marine monitoring,resource exploration,crop estimation,mapping,and military applications,and can play a role that other remote sensing methods can not play.Therefore,it is getting more and more attention from countries around the world.The polarized multi-angle SAR image is a new imaging system.The multi-angle observation system provides more angle information than the conventional strip SAR system.Based on the polarization information,it is collected in multiple angle apertures from the 360 ° range.which can obtain richer information,and hope to avoid classification errors caused by the singularity of the observation direction,which helps to polarize multiple angles.Research on Classification and Recognition of SAR Images.A large number of polarized multi-angle SAR image classification has been studied,but further research is needed.The optimized recognition algorithm for polarized multi-angle SAR images mainly includes two parts: feature engineering and classifier construction.Optimization of features: First,in order to solve the problem that the inevitable coherent speckle noise of the polarized SAR image imaging can cause the same-spectrum foreign matter,the same-object hetero-spectrum,etc.,by analyzing the sequence data curve,according to the sequence curve Features for data processing.Aiming at the problem that the experimental data is sequence data,the target decomposition of the sequence data of ten apertures is used to obtain the sequencebased feature parameters.The ten aperture sequence data of each pixel is expanded,and each pixel of the classified image is used as a classification.The target method obtains the feature data.After verification,the results show that the feature data suitable for the input of the classifier can be obtained by this method.Finally,the optimization of the classifier is based on the small sample size of the polarized multi-angle SAR data and the applicability of the classification of experimental data.How to choose a suitable classifier to classify and optimize the data,adopting the data that is suitable for small samples and is more robust A good Support Vector Machine(SVM)classification and recognition technology classifies and recognizes polarized multi-angle SAR data.The introduction of the Markov Random Field(MRF)method improves the SVM classification method,taking into account the single The classifier is not good enough to balance the problem of high-dimensional complex features.A comparative experiment was carried out through the Light GBM integrated classification method;and the applicability of SVM for polarized multi-angle SAR image classification was carried out,and the combined classification of MRF and SVM classifier and Light GBM were analyzed.Integrating the classification results,the classification accuracy of both methods has been significantly improved.