Segmentation of Remote Sensing by Fusing Grafting-Type Attention and Detail Perception

Yijie Zhang; Xinlin Xie; Jing Fan; Zeyun Duan

doi:10.3788/LOP240674

Laser & Optoelectronics Progress, Volume. 61, Issue 24, 2428009(2024)

Segmentation of Remote Sensing by Fusing Grafting-Type Attention and Detail Perception

Yijie Zhang^1,2, Xinlin Xie^1,2、*, Jing Fan^1,2, and Zeyun Duan^1,2

¹School of Electronic and Information Engineering, Taiyuan University of Science and Technology, Taiyuan 030024, Shanxi , China

²Shanxi Key Laboratory of Advanced Control and Equipment Intelligence, Taiyuan 030024, Shanxi , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(11)

Fig. 1. Framework diagram of superpixel segmentation of remote sensing by fusing grafting-type attention and detail perception

Download full size

Fig. 2. Spatial detail module for edge guidance

Download full size

Fig. 3. Structure of grafting-type attention mechanism

Download full size

Fig. 4. Texture-aware loss structure diagram

Download full size

Fig. 5. Performance comparison for different values of K

Download full size

Fig. 6. Visualization results of different algorithms

Download full size

Table 1. Performance comparison of different detailing modules
View table
Table 1. Performance comparison of different detailing modules
Detail processing module Filter UE BR ASA
— × 0.1814 0.8021 0.9455
Filter √ 0.1792 0.7592 0.9339
FRH^［26］ × 0.1757 0.8094 0.9436
ESDP- × 0.1673 0.8569 0.9479
ESDP √ 0.1586 0.8720 0.9549

Table 2. Performance comparison of different attention mechanisms
View table
Table 2. Performance comparison of different attention mechanisms
Attention mechanism UE BR ASA
— 0.1824 0.7268 0.9320
SE^［17］ 0.1764 0.7966 0.9435
DANet^［21］ 0.1764 0.7990 0.9436
FLA^［22］ 0.1703 0.8161 0.9447
GTAM 0.1586 0.8720 0.9549

Table 3. Comparison of performance using texture-aware function or not
View table
Table 3. Comparison of performance using texture-aware function or not
Texture-aware loss UE BR ASA
× 0.1752 0.7369 0.9456
√ 0.1586 0.8720 0.9549

Table 4. Comparison of small target performance
View table
Table 4. Comparison of small target performance
Algorithm UE BR ASA
SLIC 0.2653 0.6072 0.9299
SCN 0.2027 0.6378 0.9392
RIM 0.3189 0.7167 0.9061
AINET 0.1906 0.6772 0.9322
RUN 0.2086 0.6967 0.9474
Proposed algorithm 0.1670 0.7971 0.9533

Table 5. Performance comparison of different algorithms on the Potsdam dataset
View table
Table 5. Performance comparison of different algorithms on the Potsdam dataset
Algorithm UE BR ASA MDE
SLIC 0.2198 0.7289 0.9347 3.9734
SCN 0.2019 0.8297 0.9397 3.6952
RIM 0.1904 0.7766 0.9363 3.2198
AINET 0.1932 0.8333 0.9488 3.7776
RUN 0.1746 0.8641 0.9559 3.4374
Proposed algorithm 0.1586 0.8739 0.9549 2.5989

Tools

Get Citation

Copy Citation Text

Yijie Zhang, Xinlin Xie, Jing Fan, Zeyun Duan. Segmentation of Remote Sensing by Fusing Grafting-Type Attention and Detail Perception[J]. Laser & Optoelectronics Progress, 2024, 61(24): 2428009

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites