一种军事图像目标识别神经网络系统

A Military Image Target Recognition Neural Network System

在军事图像目标识别中,目标通常发生比例、平移、旋转变化,有时还处于复杂背景之中或部分被障碍物遮挡,而识别往往要求是实时的。这使得传统的图像目标识别方法不能获得较好的效果。本文提出了一种神经网络目标识别系统,该系统能直接识别图像目标,而无需提取图像中的目标特征,具有目标识别的比例、平移、旋转不变性,具有良好的复杂背景下的目标识别性能,是一种高速、实用、识别率高的军事图像目标识别神经网络系统。文中给出了改进的神经网络模型并针对不同军事目标的识别需要进行了仿真实验。

For military image target recognition, we propose a neural target recognition system which has high recognition speed and is insensitive to translation, rotation, and scale in this paper. The system consists of a fixed invariant network with appendent nodes and a trainable mutilayered network. Simulation results and analysis are given at the end of this paper. The result shows that the system approach works well for single or multiple military targets. It is promise in military image target recognition research.