基于Hopfield自联想神经网络的天幕靶信号识别技术

发布时间：2018-07-16 22:54

【摘要】：为了提高天幕靶系统测试精度和可靠性,探索测试产生各种干扰噪声如弹头激波、弹底激波、蚊虫飞鸟、振动等干扰因素的影响规律,利用Hopfield自联想神经网络的方法,识别并剔除典型因素干扰。通过对实弹射击试验得到的数据进行分析,充分验证了天幕靶系统的准确性和可靠性。分析结果表明:与电平信号识别相比,在射频为5发/min、口径为23 mm的炮弹测试中,自联想神经网络信号识别率提高了17.2%;在弹型为穿甲弹,口径为23 mm的测试中,Hopfield自联想神经网络信号识别率提高了46.7%;对于射频为7500发/min的天幕靶连发弹丸信号测试条件下,正确信号识别率均达到了93%以上。在复杂环境条件下,Hopfield神经网络算法识别率远远高于传统的电平识别,提高了信号的识别率,能够适应一定区域内的复杂环境因素。
[Abstract]:In order to improve the test accuracy and reliability of the sky screen target system, and to explore the influence rules of the interference factors such as warhead shock wave, projectile bottom shock wave, mosquito flying bird, vibration and so on, the method of hopfield autoassociative neural network is used. Identify and eliminate typical factors. The accuracy and reliability of the sky screen target system are fully verified by analyzing the data obtained from the live firing test. The analysis results show that compared with the level signal recognition, the recognition rate of the ARN signal is improved by 17.2in the test of the projectile with radio frequency of 5 / min and caliber of 23mm, and the projectile is armour-piercing bomb. The recognition rate of hopfield autoassociative neural network signal is increased by 46.7 in the test of 23 mm caliber, and the correct signal recognition rate is over 93% for the continuous projectile signal with radio frequency of 7500 beats / min. The recognition rate of hopfield neural network algorithm in complex environment is much higher than that of traditional level recognition algorithm, which improves the recognition rate of signal and adapts to complex environment factors in a certain area.
【作者单位】：长春理工大学空间光电技术研究所;西安工业大学光电工程学院;
【基金】：国家自然科学基金项目(61275080) 吉林省科技发展计划项目(20140520115JH)
【分类号】：TN911.7

【参考文献】