用于交通图像识别的改进尺度依赖池化模型

被引：5

作者：

徐喆

冯长华

机构：

[1] 北京工业大学信息学部

来源：

计算机应用 | 2018年 / 38卷 / 03期

关键词：

卷积神经网络; 交通标志识别; 尺度依赖池化; 特征融合; 空间金字塔池化;

D O I：

暂无

中图分类号：

TP391.41 [];

学科分类号：

080203 ;

摘要：

针对交通标志在自然场景中所占的比例较小、提取的特征量不足、识别准确率低的问题,提出改进的尺度依赖池化(SDP)模型用于小尺度交通图像的识别。首先,基于神经网络深卷积层具有较好的轮廓信息与类别特征,在SDP模型只提取浅卷积层特征信息的基础上,使用深卷积层特征补足型SDP(SD-SDP)映射输出,丰富特征信息;其次,因SDP算法中的单层空间金字塔池化损失边缘信息,使用多尺度滑窗池化(MSP)将特征池化到固定维度,增强小目标的边缘信息;最后,将改进的尺度依赖池化模型应用于交通标志的识别。实验结果表明,与原SDP算法比较,提取特征量增加,小尺度交通图像的识别准确率较好地提升。

引用

页码：671 / 676

页数：6

共 22 条

[1] Man vs. computer: Benchmarking machine learning algorithms for traffic sign recognition[J] . J. Stallkamp,M. Schlipsing,J. Salmen,C. Igel. &nbspNeural Networks . 2012
[2] Multi-column deep neural network for traffic sign classification[J] . &nbspNeural Networks . 2012
[3] Traffic sign recognition using group sparse coding[J] . Huaping Liu,Yulong Liu,Fuchun Sun. &nbspInformation Sciences . 2014
[4] Traffic sign detection via interest region extraction[J] . Samuele Salti,Alioscia Petrelli,Federico Tombari,Nicola Fioraio,Luigi Di Stefano. &nbspPattern Recognition . 2014
[5] 基于双三次插值算法的图像缩放引擎的设计
张阿珍
刘政林
邹雪城
向祖权
[J]. 微电子学与计算机, 2007, (01) : 49 - 51
[6] Traffic sign recognition based on attribute-refinement cascaded convolutional neural networks. XIE K,GE S,YE Q,et al. Proceedings of the 17th Pacific-Rim Conference on Multimedia,LNCS 9916 . 2016
[7] Toward an optimal convolutional neural network for traffic sign recognition. AGHDAM H H,HERAVI E J,PUIG D. Proceedings of the 8th International Conference on Machine Vision . 2015
[8] Cascaded neural network with scale dependent pooling for object detection. CHOI W,YANG F,LIN Y. U.S.Patent Application15/343,017 . 2016
[9] Very deep convolutional networks for large-scale image recognition. SIMONYAN K,ZISSERMAN A. Computer science . 2014
[10] Real-time traffic sign recognition from video by class-specific discriminative features[J] . Andrzej Ruta,Yongmin Li,Xiaohui Liu. &nbspPattern Recognition . 2009 (1)

← 1 2 3 →