The Study Of Radar Target Feature Extraction And Recognition

The paper selected topic originates from the national defense science and industry committee advanced research project "the radar target RCS survey research" the sub- topic "the radar target feature extraction and recognition". Entrusts with the radar system by the target identification function is the direction which the modern radar develops, the our country active duty majority radar all only can gain the goal the glint information, therefore, how studies carries on the target identification using the goal glint information to have the important practical significance.The radar target identification mainly contains two aspects the duties: First is the characteristic withdraws; Next is the target category. This article take the modern signal processing technology as a foundation, proposed withdraws the method based on the wavelet transformation and the double spectrum radar target characteristic; The preceding method is uses the signal in the frequency range energy distribution trick recognition goal, the actual test obtains approximately 83% average recognition rate. The latter method uses the statistical distribution stable spectrum coefficient target discrimination, actually tests the average recognition rate 78%. Next, the paper in detail discussed the SOM nerve network structure and the characteristic, through the improvement training algorithm, has overcome the SOM nerve network inherent training result unstable flaw, will improve the method practical application to the radar target identification sorter design, the actual test will obtain 75% average recognition rate.