Based On Support Vector Machine (svm) Of Synthetic Aperture Radar Images Target Recognition

As a representative of microwave remote sensing, Synthetic aperture radar (SAR) has a unique advantage in the field of remote sensing on earth sciences. SAR has been widely applied to military, agriculture, geology, marine monitoring, disaster monitoring and other fields. The field of image processing and interpreting has been focused on achieving SAR image target recognition. As a new generation of machine learning techniques Based on statistical learning theory, Support Vector Machine (SVM) training in small samples, non-linear case has good generalization performance.This thesis studies SAR image target recognition based on SVM methods. The main contributions are as follows:Firstly, introduces SVM theory, and sums up the basic idea of SVM and application directions. Secondly, reviews SAR image target research status, sums up the SAR image target recognition processes and key technologies. Thirdly, introduces feature extraction method based on wavelet transform and verify that has high accuracy. And introduces kernel principal component analysis method (KPCA), the kernel mode avoids complicated dot product operation, solve tremendous computation problems in large-scale sample, and improve the operation speed; Fourthly, Make incremental training algorithm, that is according to the training to support distribution of vector, using support vector instead of sample set, enabling small matrix calculations. In simulation experiments, SAR image target recognition based on incremental algorithm of support vector machine has high recognition rate in small samples, and increases speed, reduce training time.Finally, this thesis uses KPCA feature selection method and the method of support vector machines for SAR image Target recognition. Experimental results shows:KPCA feature selection algorithm and SVM incremental training method can obtain a higher recognition accuracy and shortly run time,this is an effective method for SAR target recognition.