基于BP神经网络的汽车车型识别方法

将BP神经网络应用于汽车车型的自动识别,在简述车型图像预处理和特征抽取的基础上,重点讨论学习算法的选取问题,提出一种分阶段的训练方法。对三类车型的36对样本做训练和测试实验,实验结果表明,所采用的训练算法优于传统BP算法和共轭梯度法,所设计的网络能快速有效地识别汽车车型。

A Recognition of Automobile Types Method Based on the BP Neural Network

This paper applys BP neural network into the auto-matic recognition of automobile models.On the basis of the pretreatment of the images of the automobile type and character abstraction,it mainly set forth how to choose the learning algo-rithm and put forward a phase learning method.The result of experiments on36samples of three kinds of automobile models indicates that the train method is superior to the BP algorithm and conjugate gradient method and the designed can recognize automobile models rapidly and effectively.