New Method to Improve Classification of Radar Object by Different Kernel of Support Vector Machine

Today, the importance of classification of radar object has drastically grown and numerous methods have been applied to achieve this goal. Support Vector Machine (SVM) stands among the newest methods on this subject. Herein, three different types of SVM methods have been suggested for fighter, airplane and helicopter including on-vs-one method, one-vs-rest method and directional acyclic graph method. Since these methods were not sufficiently capable of being distinctive in a linear way, some concepts of Kernel function such as polynomial, linear, quadratic and basic radial function have been used. Directional acyclic graph method using Kernel function yielded the best results according to the outputs obtained from simulation. One-vs-rest method using RBF and quadratic Kernel as well, was more adapted than on-vs-one method. The run time of performing these three methods is also deeply verified. The results showed a similar run time for all the three. Hence, to classify the noted goal, the method of directional acyclic graph is proposed as it manifests the most optimized performance in terms of accuracy.