本文利用BP(Back-Propagation)人工神经网络对三维物体的姿态测定进行了研究。姿态测定一直缺少通用而实际的方法,人工神经网络由于具有强大的自组织、自适应学习能力,迅速的并行信息处理能力,可望解决这个问题。但现有BP算法存在训练慢和易陷入局部最小两个问题.本文提出的级联形式网络结构,使BP网络的训练速度大为提高,陷入局部最小的可能性大为降低。利用这种级联结构对飞机模型姿态测定,取得了较好的实验结果。 In this paper, BP (Back-Propagation) artificial neural network for three-dimensional object attitude determination. Attitude measurement has always been a lack of general and practical methods, artificial neural network with a strong self-organization, adaptive learning ability, rapid parallel information processing capabilities, is expected to solve this problem. However, the existing BP algorithm has two problems: slow training and easy to fall into local minima.This paper proposes a cascade network structure, which greatly improves the training speed of BP network and greatly reduces the possibility of falling into a local minimum. Using this cascade structure of aircraft model attitude determination, and achieved good experimental results.