Share
تمييز الاهداف الرادارية وذلك عن طريق استخدام الشبكات العصبية وافضلية الحشد الجزيئي == Radar Target Recognition Using Neural Network And Particle Swarm Optimization
Author name:
علي ناظم غالب
Supervisor name:
وليد خالد عبد علي
General topic:
Electrical, Electronic and Communications Engineering
Specific topic:
Electronics and Communications Engineering
Degree:
Master
University:
Mustansiriyah University - College Of Engineering - Department Of Electrical Engineering
Language:
English
University location:
Baghdad
First pages:
34T535 - p.pdf
Abstract:
In the last decades, the pattern recognition and classification are taking an important area of the research. One of the main applications of pattern recognition is the radar target recognition which is vital in radar system.In this this work, a radar signal were used and analyzed for three targets present a contribution is this trend (aircrafts SR - 71, C - 130 and AH - 64), and used as initial data, then variations are made to simulate the target signal variation (50 range profile for each target).A wavelet transform (multi resolution decomposition) is used to reduce the dimensions of radar range profile and then leads to reduce storage and computation.The wavelet transform including two procedures, the first one using approximation and details coefficients as feature extraction, while second procedure use related with the details coefficient only.Artificial Neural Network (ANN) is used as recognizer. The Feed Forward Neural Network (FFNN) and Elman Recurrent Neural Network (ERNN) are used. The network was trained using Supervised Learning Algorithms. The FFNN was trained by (Levenberg - Marquardt (lm), Bayesian regulation (br), particle swarm optimization). The (ERNN) was trained by (lm, br).The best recognition rate of three targets is obtained by using the feature extraction from (normalized details wavelet coefficient) and the ERNN trained by (br). The recognition rate in this procedure is about 96% compare with other methods.When adding the AWGN S/N=5db, the recognition rate reducing to 85%.
