用于跟踪机动和非机动目标的算法有很多[1]—[4]。对机动目标和非机动目标可以使用不同的算法。检测目标机动的能力能为给定的态势提供良好的算法。通常采用一些特定的技术实现机动检测。神经网络为实现目标机动检测提供了一种新工具。已经为这种应用开发测试了两个不同的网络。一个网络使用方位数据,另一个网络使用滤波器残差。方位数据先被仿真,然后被滤波,以便为这两种网络提供所需要的输入。使用反向传播算法训练每个网络,该算法是一种被资料证明的普通梯度下降训练算法[5]。已经研究了很多不同的多层感知结构。这里介绍的两种结构能在训练和测试时提供最好的性能。对这两个网络进行了相互比较,并且与一个特定的门限算法进行了比较,给出了相对性能数据。这项研究结果表明,利用神经网络实现目标机动检测是可能的。 There are many algorithms for tracking maneuvering and non-maneuvering targets [1] - [4]. Different algorithms can be used for maneuvering targets and non-maneuvering targets. The ability to detect a target maneuver can provide a good algorithm for a given situation. Usually use some specific technology to achieve motor detection. Neural networks provide a new tool for achieving target maneuver detection. Two different networks have been tested for this application development. One network uses azimuth data and the other network uses filter residuals. Orientation data is first simulated and then filtered to provide the required input for both networks. Each network is trained using a back propagation algorithm, which is a data-proven, normal gradient descent training algorithm [5]. Many different multilayer sensing structures have been studied. The two structures presented here provide the best performance during training and testing. The two networks are compared with each other and compared with a particular threshold algorithm to give relative performance data. The results of this study show that it is possible to use neural networks to achieve the target maneuver detection.