基于神经网络的地形匹配方法研究

Terrain matching based on neural network

提出了基于神经网络实现多特征融合的地形匹配算法,充分利用地形的各种不同的统计特征和几何特征,构造了一种地形匹配网络模型.通过对实时图和基准图的分析,给出了计算网络节点之间的权值函数,建立了网络系统能量方程,通过求系统的最小能量得到最佳匹配位置.由于网络能融合地形的不同统计特征和几何特征,所以算法大大提高了系统的抗干扰能力和定位精度,适合于实时图容易发生畸变的地形匹配领域.实验结果表明,定位精度和抗干扰能力均优于传统的地形匹配方法.

A new terrain matching neural network algorithm mode is constructed by means of multifeature fusion, which includes different statistical and geometrical features. By analyzing of the real terrain and the reference terrain, the connective weight function between different nodes is deduced, and the energy formula of the network system is structured. The matching position can be acquired by seeking the least system energy. As the algorithm can utilize different features sufficiently, it has higher matching precision, and also has better robustness to noise and geometric distortion, The ex- perimental results reveal that its performance is better than that by the conventional terrain matching modes.