基于小波变换的汽车车型识别

Vehicle Model Recognition Based on Wavelet Transform

介绍了多品种混流机器人喷漆自动线的汽车车型识别系统的一种识别方法。主要是利用小波变换具有良好的多尺度特征表达能力 ,以及能在空域、频域和方向上分别进行分解 ,而且能够去除冗余信息和噪声等优点对汽车图像进行分解。然后把分解后的图像作为多层前馈神经网络的输入节点 ,对自动喷漆线上的汽车车型进行识别。结果表明 :基于小波变换的神经网络汽车车型识别方法 ,能够识别返修车 。

This paper introduces a recognition method on different models of vehicle on a multi-model mixed robot painting line.Taking the advantage of wavelet transform in expressing multi-scale features,vehicle images are decomposed in spatial,frequency and orientation domains respectively with redundant information and noise removed. The decomposed images are then used as the nodes on input-layer of multi-layer feed-forward network,which recognize the vehicle model on painting line.The results show that the method based on wavelet transform enhances the accuracy of recognition on vehicle model.