BP在传感器数据融合中的应用.docVIP

浙 江 理 工 大 学 信 息 电 子 学 院 毕业论文(设计)诚信声明 我谨在此保证：本人所写的毕业论文(设计)，凡引用他人的研究成果均已在参考文献或注释中列出。论文(设计)主体均由本人独立完成，没有抄袭、剽窃他人已经发表或未发表的研究成果行为。如出现以上违反知识产权的情况，本人愿意承担相应的责任。 声明人(签名)： 年 月 日 摘 要 现实应用中，传感器静态特性常常受一个或多个非目标参量的影响，导致性能不稳定，测量准确度低。为了提高传感器的稳定性，可采用有效的融合处理方法来消除非目标参量对传感器输入-输出特性的影响。数据融合的算法很多，常用的有Bayes决策理论、卡尔曼滤波法、模糊融合、BP神经网络融合等。 BP神经网络是模拟人工智能的一条途径，特别是可以利用BP神经网络解决人工智能研究中所遇到的一些难题。在模拟与数字混合的超大规模集成电路制作技术提高到新的水平，完全付诸实用化背景下经网络寻求出路的时机已经成熟。权值误差改变很小，使训练过程几乎停顿关键词 Abstract In real applications, the static characteristics of sensor will be influenced by one or several non-target parameters, which results in unstable performance and low accurate measurement. In order to improve stability of sensor, the data fusion can be an effective approach to eliminate the influence of non-target parameters on the sensor input-output characteristic. There are a lot of data fusion algorithms, commonly used with Bayes decision theory, Kalman filter method, fuzzy fusion, BP neural network integration. In particular, BP neural networks can solve some problems that artificial intelligence can not do. Manufacturing technology of mixed analog and digital VLSI has developed a new level, completely put to practical use. Against this background, the time of artificial neural network to find a way out is ripe. Therefore, this paper studies the application of BP neural network processing in sensor data fusion. In this paper, I adopt the BP neural network to data fusion, to achieve the BP network algorithm through the C programming language. Programming environment is based on VC++, multi-sensor data fusion can be achieved through this program, thus reaching improvement of sensor measurement accuracy. Then, it will be verified through Matlab simulation, at last, using different training algorithms study performance of BP algorithm. BP algorithm has disadvantage of slow convergence, and in the latter training, weigh error is very small, halting the training process, which makes the neural network trai