• Optical Instruments
  • Vol. 42, Issue 4, 33 (2020)
Jianpeng SU, Yingping HUANG*, Bogan ZHAO, and Xing HU
Author Affiliations
  • School of Optical-Electrical and Computer Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China
  • show less
    DOI: 10.3969/j.issn.1005-5630.2020.04.006 Cite this Article
    Jianpeng SU, Yingping HUANG, Bogan ZHAO, Xing HU. Research on visual odometry using deep convolution neural network[J]. Optical Instruments, 2020, 42(4): 33 Copy Citation Text show less
    Flow diagram of visual odometry end-to-end network
    Fig. 1. Flow diagram of visual odometry end-to-end network
    Inception structure
    Fig. 2. Inception structure
    The trajectory estimation of 04,05,07 and 10 sequence
    Fig. 3. The trajectory estimation of 04,05,07 and 10 sequence
    层数卷积核大小步长通道数
    卷积层17×7264
    卷积层25×52128
    卷积层35×52256
    卷积层43×31256
    卷积层53×32512
    卷积层63×31512
    卷积层73×32512
    卷积层83×31512
    卷积层93×321024
    Table 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15×52128
    卷积层23×32512
    卷积层33×31512
    卷积层43×32512
    卷积层53×31512
    卷积层63×321024
    Table 2.

    Deep-CNN-VO subconvolution parameter

    Deep-CNN-VO子卷积参数

    序列Deep-CNN-VOCNN-VOVISO2-M[4]DeepVO[10]
    ATE/%ARE/((°)/m)ATE/%ARE/((°)/m)ATE/%ARE/((°)/m)ATE/%ARE/((°)/m)
    038.790.046715.530.06528.470.08828.490.0689
    0411.870.068218.050.02584.690.04497.190.0697
    056.980.03717.920.052919.220.03542.620.0361
    079.810.093313.940.083123.610.04113.910.0460
    1017.750.045821.450.105941.560.32998.110.0883
    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

    Jianpeng SU, Yingping HUANG, Bogan ZHAO, Xing HU. Research on visual odometry using deep convolution neural network[J]. Optical Instruments, 2020, 42(4): 33
    Download Citation