• Laser & Optoelectronics Progress
  • Vol. 59, Issue 12, 1228002 (2022)
Rongping Zou1、2, Bin Zhu1、2、*, Chenyang Wang1, Yaoxuan Zhu1、2, and Yangdi Hu3
Author Affiliations
  • 1College of Electronic Engineering, National University of Defense Technology, Hefei 230037, Anhui , China
  • 2Key Laboratory of Infrared and Low Temperature Plasma of Anhui Province, Hefei 230037, Anhui , China
  • 3Army 32256 of PLA, Guiling541000, Guangxi , China
  • show less
    DOI: 10.3788/LOP202259.1228002 Cite this Article Set citation alerts
    Rongping Zou, Bin Zhu, Chenyang Wang, Yaoxuan Zhu, Yangdi Hu. Heterogeneous Remote Sensing Image Matching Algorithm Based on Residual Pseudo-Siamese Convolution Cross-Correlation Network[J]. Laser & Optoelectronics Progress, 2022, 59(12): 1228002 Copy Citation Text show less
    Structure of residual pseudo-siamese convolution cross-correlation network
    Fig. 1. Structure of residual pseudo-siamese convolution cross-correlation network
    Heat map of inference results for different networks in different terrains
    Fig. 2. Heat map of inference results for different networks in different terrains
    Heat map of middle layer features in RPSCCNet and heat map of result before Softmax operation for two networks
    Fig. 3. Heat map of middle layer features in RPSCCNet and heat map of result before Softmax operation for two networks
    TerrainEvaluating indicatorPSiamNet(c)Adapted HardNet(d)CorASLNet(e)RPSCCNet(f)
    ForestHeatmap quality2862.8811939.880.2728.14
    (1)L2 distance /pixel1.006.080.002.00
    Urban AreaHeatmap quality11379.4013830.627.0811.78
    (2)L2 distance /pixel3.004.240.000.00
    FarmlandHeatmap quality17970.9415000.807360.9927.42
    (3)L2 distance /pixel4.475.0045.692.00
    RiverHeatmap quality18175.439731.151813.5632.71
    (4)L2 distance /pixel51.2478.8164.282.00
    Water SurfaceHeatmap quality11373.0614464.50136.0641.58
    (5)L2 distance /pixel80.6066.2467.674.47
    Gobi(flat)Heatmap quality17252.409498.53474.1354.71
    (6)L2 distance /pixel45.1832.5769.0753.66
    GobiHeatmap quality16796.477643.4327.0127.87
    (7)L2 distance /pixel7.0787.692.002.00
    Hilly AreaHeatmap quality13250.7416251.0635.372.66
    (8)L2 distance /pixel3.1666.6710.000.00
    River valleyHeatmap quality4118.5312901.604.731.53
    (9)L2 distance /pixel3.008.600.00.00

    Highland

    (10)

    Heatmap quality9106.8816033.428.694.92
    L2 distance /pixel1.0029.830.000.00
    HighlandHeatmap quality11021.9712959.8930.1733.03
    (11)L2 distance /pixel9.2163.972.002.00
    HighlandHeatmap quality7603.9716355.2868.421.44
    (12)L2 distance /pixel5.0086.3784.850.00
    HighlandHeatmap quality8796.7218123.400.392.86
    (13)L2 distance /pixel2.2369.770.000.00
    HighlandHeatmap quality13597.9912910.650.934.24
    (14)L2 distance /pixel3.6083.630.000.00
    HighlandHeatmap quality5318.5015361.870.482.00
    (15)L2 distance /pixel70.1774.810.000.00
    Mean Value of Heatmap quality11247.0013532.41177.8818.46
    Mean Value of L2 distance /pixel19.3350.9523.034.54
    Median Value of Heatmap quality11373.4913830.6227.0111.78
    Median Value of L2 distance /pixel4.4766.242.000.00
    Table 1. Quality of corresponding normalized heat map and L2 distance error of inference result in Fig. 2
    Evaluating indicatorRPSCCNetCorASLNetPSiamNetAdapted HardNet
    Mean value of L2 distance /pixel5.34712.2717.8635.03
    ACC(L2≤3 pixel)/%91.0379.6555.3216.72
    ACC(L2≤2 pixel)/%82.4974.5323.046.01
    ACC(L2≤1 pixel)/%55.4952.476.501.75
    ACC(L2=0 pixel)/%17.3717.020.000.00
    Table 2. Matching accuracy and mean value of L2 distance of four methods in test set
    ParameterRPSCCNetCorASLNetPSiamNetAdapted HardNet
    Size of Network /MB85.404.9641.005.16
    FPS(Network)82.88183.37165.46364.82
    FPS(Single match)82.88183.370.010.02
    Table 3. Performance comparison of four methods in Nvidia RTX 2080Ti
    Rongping Zou, Bin Zhu, Chenyang Wang, Yaoxuan Zhu, Yangdi Hu. Heterogeneous Remote Sensing Image Matching Algorithm Based on Residual Pseudo-Siamese Convolution Cross-Correlation Network[J]. Laser & Optoelectronics Progress, 2022, 59(12): 1228002
    Download Citation