Accurate arrested landing state recognition of carrier-based aircraft based on coordinate attention and weighted bi-directional feature pyramid network

Zhe LI; Jie YANG; Yi ZHANG; Hua WANG; Yafei LI; Ke WANG; Mingliang XU

doi:10.19693/j.issn.1673-3185.04005

Chinese Journal of Ship Research, Volume. 20, Issue 4, 124(2025)

Accurate arrested landing state recognition of carrier-based aircraft based on coordinate attention and weighted bi-directional feature pyramid network

Zhe LI¹, Jie YANG¹, Yi ZHANG², Hua WANG^1,3,4, Yafei LI^1,3,4, Ke WANG^1,3,4, and Mingliang XU^1,3,4

Author Affiliations

¹School of Computer and Artificial Intelligence, Zhengzhou University, Zhengzhou 450001, China

²Unit 96813 of PLA, Huangshan 245000, China

³National Supercomputing Center in Zhengzhou, Zhengzhou 450001, China

⁴Engineering Research Center of Intelligent Swarm Systems, Ministry of Education, Zhengzhou 450001, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(17)

Fig. 1. [in Chinese]

Fig. 2. [in Chinese]

Fig. 3. [in Chinese]

Fig. 4. [in Chinese]

Fig. 5. [in Chinese]

Fig. 6. [in Chinese]

Fig. 7. [in Chinese]

Fig. 8. [in Chinese]

Fig. 9. [in Chinese]

Fig. 10. [in Chinese]

Download full size

View in Article

Table 1. Experimental software and hardware environment
View table
View in Article
Table 1. Experimental software and hardware environment
软件环境和硬件配置参数
操作系统 Windows 10 21H2
CPU I5-9300H
GPU配置 GTX-1650-4G
CUDA 11.3（CuDNN 8.2）
算法环境 TensorRT 8.2 Pytorch 1.11.0
Opencv 4.5.3 PyQt5.12

Table 2. Comparison of model performance parameters when adding CA to different positions
View table
View in Article
Table 2. Comparison of model performance parameters when adding CA to different positions
模型精度/% mAP₅₀/%
YOLOv5s 78.9 74.8
CA_13 81.7 75.9
CA_17 80.2 73.3
CA_20 81.2 72.9

Table 3. Comparison of model performance parameters after replacing C2F at different positions
View table
View in Article
Table 3. Comparison of model performance parameters after replacing C2F at different positions
模型精度/% mAP₅₀/%
YOLOv5s 78.9 74.8
更换全部C3结构 82.2 72.6
更换Backbone的C3结构 76.2 72.7
更换Head的C3结构 80.1 75.6

Table 4. Comparison of experimental results in the CATHR-DET

View table

View in Article

Table 4. Comparison of experimental results in the CATHR-DET

方法	精度/%	召回率/%	F1	mAP₅₀/%	FPS/(帧·s⁻¹)
YOLOv5s	83.3	71.7	77.1	78.6	53.8
SSD	77.8	33.2	46.5	56.6	55.2
Faster-RCNN	80.9	46.4	58.9	77.5	29.6
RetinaNet	67.7	79.5	73.1	65.2	33.2
CenterNet	73.3	43.4	54.5	67.6	37.1
本文	86.8	80.9	83.6	82.9	56.5

Table 5. Comparison of experimental results in the HCESI-DET

View table

View in Article

Table 5. Comparison of experimental results in the HCESI-DET

方法	精度/%	召回率/%	F1	mAP₅₀/%	FPS/(帧·s⁻¹)
YOLOv5s	78.9	74.8	73.4	75.4	52.6
SSD	72.4	46.4	56.6	51.7	54.2
Faster-RCNN	76.8	49.2	56.5	70.5	27.4
RetinaNet	67.7	77.5	72.3	63.6	32.1
CenterNet	71.2	41.3	52.3	65.2	35.8
本文	79.5	74.2	76.8	77.9	55.2

Table 6. Comparison of experimental results in the VOC dataset
View table
View in Article
Table 6. Comparison of experimental results in the VOC dataset
方法精度/% 召回率/% F1 mAP₅₀/%
YOLOv5s 72.4 66.2 69.1 71.9
SSD 66.3 38.4 48.6 69.7
Faster-RCNN 75.8 53.7 62.8 65.7
RetinaNet 71.9 60.5 65.7 71.4
CenterNet 65.8 33.9 44.7 66.4
本文 80.9 76.4 78.6 79.4

Table 7. Comparison of ablation experiment results
View table
View in Article
Table 7. Comparison of ablation experiment results
方法 CA BiFPN C2F mAP₅₀/%
YOLOv5s — — — 74.8
YOLOv5s+CA √ — — 75.9
YOLOv5s+BiFPN — √ — 76.1
YOLOv5s+C2F — — √ 75.6
本文 √ √ √ 77.9

Tools

Get Citation

Copy Citation Text

Zhe LI, Jie YANG, Yi ZHANG, Hua WANG, Yafei LI, Ke WANG, Mingliang XU. Accurate arrested landing state recognition of carrier-based aircraft based on coordinate attention and weighted bi-directional feature pyramid network[J]. Chinese Journal of Ship Research, 2025, 20(4): 124

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites