论文部分内容阅读
开发了用于毫米波(MMW)雷达数据自动目标识别(ATR)系统的神经网络分类机。采用测绘模型(后向传播和相对传播)开发两种不同的神经网络分类机,并与二次(类似贝叶斯)分类机作了比较。采用统计特征值集合和雷达数据特征值集合对所有三种分类机系统进行了测试和试验。采用统计特征值集合试验MMW ATR一般优先于采用实际数据。试验结果表明,后向传播网络对随机特征值集合的精度接近100％,稍好于相对传播模型效果。
A neural network sorter for the automatic target recognition (ATR) system of millimeter-wave (MMW) radar has been developed. Two different neural network classifiers were developed using mapping models (backward propagation and relative propagation) and compared with quadratic (similar Bayesian) classifiers. All three sorter systems were tested and tested using a set of statistical eigenvalues and radar eigenvalues. The use of statistical eigenvalues to test the MMW ATR generally takes precedence over the use of actual data. The experimental results show that the accuracy of the back propagation network for the random eigenvalue set is close to 100%, slightly better than that of the relative propagation model.