Optical Instruments, Volume. 42, Issue 4, 33(2020)
Research on visual odometry using deep convolution neural network
Figures & Tables(6)
Table 1.
CNN-VO subconvolution parameter
CNN-VO子卷积参数
View tableView in ArticleTable 1.
CNN-VO subconvolution parameter
CNN-VO子卷积参数
层数 卷积核大小 步长 通道数 卷积层1 7×7 2 64 卷积层2 5×5 2 128 卷积层3 5×5 2 256 卷积层4 3×3 1 256 卷积层5 3×3 2 512 卷积层6 3×3 1 512 卷积层7 3×3 2 512 卷积层8 3×3 1 512 卷积层9 3×3 2 1024
Table 2.
Deep-CNN-VO subconvolution parameter
Deep-CNN-VO子卷积参数
View tableView in ArticleTable 2.
Deep-CNN-VO subconvolution parameter
Deep-CNN-VO子卷积参数
层数 卷积核大小 步长 通道数 卷积层1 5×5 2 128 卷积层2 3×3 2 512 卷积层3 3×3 1 512 卷积层4 3×3 2 512 卷积层5 3×3 1 512 卷积层6 3×3 2 1024
Table 3.
The comparison experimental results of 03, 04, 05, 07 and 10 test sequence
测试序列03、04、05、07、10实验结果对比
View tableView in ArticleTable 3.
The comparison experimental results of 03, 04, 05, 07 and 10 test sequence
测试序列03、04、05、07、10实验结果对比
序列 Deep-CNN-VO CNN-VO VISO2-M[ 4 ]DeepVO[ 10 ]ATE/% ARE/（（°）/m） ATE/% ARE/（（°）/m） ATE/% ARE/（（°）/m） ATE/% ARE/（（°）/m） 03 8.79 0.0467 15.53 0.0652 8.47 0.0882 8.49 0.0689 04 11.87 0.0682 18.05 0.0258 4.69 0.0449 7.19 0.0697 05 6.98 0.0371 7.92 0.0529 19.22 0.0354 2.62 0.0361 07 9.81 0.0933 13.94 0.0831 23.61 0.0411 3.91 0.0460 10 17.75 0.0458 21.45 0.1059 41.56 0.3299 8.11 0.0883
Tools
Get Citation
Copy Citation Text
Jianpeng SU, Yingping HUANG, Bogan ZHAO, Xing HU. Research on visual odometry using deep convolution neural network[J]. Optical Instruments, 2020, 42(4): 33
Paper Information
Category: DESIGN AND RESEARCH
Received: Oct. 24, 2019
Accepted: --
Published Online: Jan. 6, 2021
The Author Email: HUANG Yingping (huangyingping@usst.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20