Remote Sensing Image-Matching Network Based on Multiscale Feature Fusion and Importance Ranking Loss

Fig. 4. Flowchart of channel-space attention module, $\otimes$ denotes element-by-element multiplication, $\oplus$ denotes summation operation, $σ$ denotes Sigmoid activation function

Download full size

Fig. 5. Feature extraction network

Download full size

Fig. 6. Examples of an optical image of a remote sensing matched dataset. (a) Potsdam; (b) Toronto; (c) Vaihingen

Download full size

Fig. 7. Visualization of key points detection by three feature detectors. (a) SIFT; (b) RF-Net; (c) DFL-Net

Download full size

Fig. 8. Visualization of matching results. (a) SIFT; (b) RF-Net; (c) DFL-Net

Download full size

Table 1. Matching results of different algorithms on remote sensing image dataset
View table
Table 1. Matching results of different algorithms on remote sensing image dataset
Method NN NNT NNR MEAN
SIFT 0.622 0.622 0.846 0.696
ORB 0.687 0.687 0.894 0.756
RF-Net 0.799 0.834 0.846 0.826
CS-Net 0.791 0.825 0.848 0.822
DFL-Net 0.817 0.841 0.870 0.843

Table 2. Parameters and FLOPs of different networks
View table
Table 2. Parameters and FLOPs of different networks
Network Layers Input Resolution Parameters /10³ FLOPs /10⁹
RF-Det 10 320×240 21.89 1.69
CS-Det 10 320×240 37.12 2.85
DFL-Det 10 320×240 37.05 2.85
DFL-Des 7 512×32×32 1.33 20.13

Table 3. Results of ablation experiments for channel-space attention, SIR-loss, and DIR-loss
View table
Table 3. Results of ablation experiments for channel-space attention, SIR-loss, and DIR-loss
CBAM SIR-loss & DIR-loss NN NNT NNR MEAN
× × 0.791 0.825 0.848 0.822
√ × 0.810 0.842 0.856 0.836
× √ 0.800 0.834 0.855 0.830
√ √ 0.817 0.841 0.870 0.843

Tools

Get Citation

Copy Citation Text

Peng Chen, Beiyuan Bao, Xu Chen. Remote Sensing Image-Matching Network Based on Multiscale Feature Fusion and Importance Ranking Loss[J]. Laser & Optoelectronics Progress, 2024, 61(18): 1828003

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Remote Sensing and Sensors

Received: Dec. 28, 2023

Accepted: Feb. 5, 2024

Published Online: Sep. 14, 2024

The Author Email: Peng Chen (holmes83@163.com)

DOI:10.3788/LOP232776

CSTR:32186.14.LOP232776

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Matching results of different algorithms on remote sensing image dataset

Table 1. Matching results of different algorithms on remote sensing image dataset

Table 2. Parameters and FLOPs of different networks

Table 2. Parameters and FLOPs of different networks

Table 3. Results of ablation experiments for channel-space attention, SIR-loss, and DIR-loss

Table 3. Results of ablation experiments for channel-space attention, SIR-loss, and DIR-loss