基于实神经网络代数算法的三维测向定位算法  

A Three-dimensional Direction Location Algorithm Based on the Algebra Algorithm of Real Neural Networks

作　　者：刘李楠[1] 赵晓萌[2] 

机构地区：[1]海军蚌埠士官学校,安徽蚌埠233012 [2]安徽科技学院理学院,安徽滁州233100

出　　处：《常熟理工学院学报》2013年第2期114-117,共4页Journal of Changshu Institute of Technology

基　　金：国家自然科学基金"低频日像仪及其关键技术研究"(11173010);安徽省级高校自然科学基金"基于神经网络的蜂窝无线定位应用研究"(KJ2012Z056);安徽科技学院青年科研基金"BP神经网络在无源定位中的应用研究"(ZRC2011270)

摘　　要：目前三维空间测向定位多采用先在二维体系中计算然后转换为空间坐标的方法,正北偏差、投影变换的方位变形等会对定位精度造成很大的影响.本文立足于三维空间,利用多个测向平面形成的空间目标观测方程组,并利用实神经网络代数算法收敛速度快、时间效率高以及隐层神经元个数选取等方面的优势,构建了三站测向定位模型.仿真实验表明,本算法定位精度明显提高,在不同测向精度条件下定位误差接近CRLB,使得该算法适用于复杂的测向环境,具有很强的实用性.At present, the three-dimensional spatial location is usually first calculated in two dimensional sys-tems, and then converted to the spatial coordinate. But north deviation azimuth deformation and projection trans-formation have a great effect on positioning precision. This paper uses the space targets observation equations formed with a plurality of direction finding plane, and then uses algebra algorithm of real neural networks with its fast convergence speed, high time efficiency and the advantage of number selection of neurons in hidden lay- er. So it is first applied to the 3D direction-finding, constructing direction finding and location model based on the three-dimensional space. Simulation results show that this algorithm positioning accuracy is obviously im-proved and that positioning error is close to CRLB under the condition of different direction finding precision, so the algorithm is applicable to complex measurement environment.

关 键 词：实神经网络 三维空间 测向定位 

分 类 号：TN955[电子电信—信号与信息处理]