摘要在国家政策的指引下,国内公路建设得到了飞速发展。而由此带来的路面检测工作也变得十分繁重;另一方面,随着自主研发汽车制造业的崛起,路面检测与识别工作对于实现汽车的舒适性与安全性能有着极其重要的作用。本文在国内外对路面识别技术研究的基础上,对基于神经网络和动载的路面识别进行了深入的研究。论文的主要内容和研究成果如下:51277
1.提出了通过RBF神经网络模拟来实现对路面的识别的技术,RBF与BP网络对比之下,有着结构简洁,学习速度更快的特点。
2.提出了通过PNN神经网络模拟来实现对路面的识别的技术,PNN神经网络收敛速度快,从而非常适用于实时处理,可以完成任意的非线性变换,具有很强的容错性。
3.通过改进的粒子群优化算法,对识别结果进行优化处理。
4.设计了RBF和PNN神经网络的Matlab模拟路面识别程序。程序设计流程包括:提取特征值,构造神经网络结构,产生人工训练样本,通过人工训练样本对遵循WPSO优化算法的神经网络进行训练和验证。
通过对路面模拟数据的处理,结果表明PNN神经网络的结果为92%,比RBF神经网络的52%识别率高出许多,由此得出结论,PNN神经网络对函数的逼近要优于RBF神经网络,可以获得较优解。
毕业论文关键词:神经网络;动载;WCPSO算法;路面识别
Abstract
Under the guidance of national policy, national highway construction has been rapid development. The resulting pavement testing work has become very onerous; hand, with the rise of independent research and the automotive industry, road detection and identification work for the realization of automotive comfort and safety performance has an extremely important role. This recognition technology on the road at home and abroad on the basis of the study, based on neural network and dynamic load of the road identification conducted in-depth research. The main contents and results are as follows:
1 .Raised through the RBF neural network simulation to achieve recognition technology on the road, RBF and BP network contrast, has a simple structure, the characteristics of learning faster.
2. Be proposed through the PNN neural network simulation to achieve recognition technology on the road, PNN neural network convergence speed, making it ideal for real-time processing, you can complete an arbitrary linear transformation, and highly fault tolerant.
3 .Through improved particle swarm optimization algorithm, the recognition result is optimized.
4. Design of the RBF and PNN neural network simulation of Matlab road identification procedures. Program design process comprising: extracting characteristic values, construct neural network structure, produce artificial training samples, followed by manual training samples WPSO optimization algorithm for neural network training and validation.
By road analog data processing, the results show that neural network PNN result was 92%, compared with 52% of RBF neural network recognition rate of many, conclude, PNN neural network function approximation is superior to RBF neural network, you can get better solution.
Keywords: neural networks; dynamic loading; WCPSO algorithm; road identification
目录
第一章 绪论 1
1.1 选题背景和意义 1
1.1.1 选题背景 1
1.1.2 选题意义 2
1.2 国内外研究现状 3
1.3 存在问题 3
1.4 论文主要研究内容和结构安排 4
第二章 汽车力学振动模型的建立及数据采集 Matlab基于神经网络和动载的路面识别:http://www.youerw.com/zidonghua/lunwen_54864.html