| 基于生成对抗网络的铁路周界行人样本生成算法 |
| |
| 引用本文： | 孔祥斌,沈志忠,陈树骏.基于生成对抗网络的铁路周界行人样本生成算法[J].铁道通信信号,2019(7):57-61. |
| |
| 作者姓名： | 孔祥斌 沈志忠 陈树骏 |
| |
| 作者单位： | 通号通信信息集团有限公司 |
| |
| 基金项目： | 中国铁路总公司科技研究开发计划项目.灾害技术项目研究——铁路综合视频智能识别测评技术及标准研究.2017T001-C |
| |
| 摘 要： | 提出了一种基于生成对抗网络(GAN)的铁路周界行人样本生成算法,解决了深度网络训练时候,行人样本缺乏的问题.该算法在像素转换生成对抗网络(Pix2Pix GAN)和行人生成对抗网络(PS-GAN)的基础上,根据实际的铁路周界环境,进行了三点改进.实验说明,该算法能够生成更逼真的行人,并且和背景的融合度更高.
|
| 关 键 词： | 生成对抗网络 铁路周界 行人样本生成 |
A GAN-based Algorithm for Generating Samples of Pedestrians in High-speed Railway Perimeter Environment |
| |
| Kong Xiangbin,Shen Zhizhong,Chen Shujun.A GAN-based Algorithm for Generating Samples of Pedestrians in High-speed Railway Perimeter Environment[J].Railway Signalling & Communication,2019(7):57-61. |
| |
| Authors: | Kong Xiangbin Shen Zhizhong Chen Shujun |
| |
| Abstract: | This paper proposes a GAN-based algorithm to generate sample data of pedestrians in perimeter environment of high-speed railway,which can be used for deep neural network training.The algorithm makes three improvements over other two algorithms,pix2pix GAN and Pedestrian GAN.based on actual perimeter environment of the railway.It has been shown in test that this algorithm can generate more vivid scene of pedestrians and better correlation with the background. |
| |
| Keywords: | GAN Railway perimeter Generation of pedestrian samples |
| 本文献已被 CNKI 维普 等数据库收录！ |
