Lightweight Infrared Small Target Detection Model Based on YOLOv10n

doi:10.12382/bgxb.2025.0146

Abstract

Abstract:

Infrared small target detection has extensive applications in military fields such as infrared guidance and tracking systems and is an important area of infrared image processing.Due to the limitations in detection equipment and the lack of inherent information about infrared small targets,the existing detection methods are difficult to meet the practical performance requirements.In order to explore a lightweight and highly accurate infrared small target detection model,a lightweight and efficient YOLOv10n infrared small target detection model (L-YOLOv10n) is designed based on YOLOv10n.The SCDown module in YOLOv10n is replaced by a lightweight spatial-channel decoupled downsampling (L-SCDown) module to enhance the key features of infrared small targets with a low computational cost.A lightweight Cross-stage partial convolution with Two Fusion layers (L-C2f) module is used to replace the C2f module,thereby enhancing the edge information of small targets and extracting the multi-scale features while reducing computational cost.To address the issues of infrared small targets with few pixels and an imbalance between foreground and background,Focal Loss and a focaler intersection-over-union (Focaler-IOU) loss function are introduced,thus allowing the model to better focus on the difficult-to-detect targets.Experimental results on the public datasets SIRST-V2 and NUDT-SIRST demonstrate that L-YOLOv10n significantly outperforms the detection-based models in both detection performance and resource consumption.The detection performance of L-YOLOv10n is slightly lower than that of Transformer-based segmentation models,but its resource consumption is significantly better than those of other models.Its generalization performance on the NUDT-SIRST dataset is also significantly higher than those of most infrared small target detection models.These results demonstrate that the proposed model strikes a balance between resource consumption and high-precision detection,demonstrating its practicality.

Key words: infrared small target detection, YOLOv10n algorithm, feature extraction, lightweight

ZHAO Zichen, XIAO Haodong, LING Huanzhang. Lightweight Infrared Small Target Detection Model Based on YOLOv10n[J]. Acta Armamentarii, 2025, 46(11): 250146-.

Figures/Tables 19

Fig.1 Network structure of L-YOLOv10n

Fig.2 Network structure of SCDown

Fig.3 Network structure of L-SCDown

Fig.4 Structure of FCS

Fig.5 Network structure of C2f

Fig.6 Network structure of shallow L-C2f

Fig.7 Statistical analysis of target distribution on SIRST-V2 and NUDT-SIRST datasets

Fig.8 Typical image samples in SIRST-V2 and NUDT-SIRST datasets,their grayscale distribution diagrams and scene

Table 1 Comparison results of different models on SIRST-V2 dataset

Fig.9 Comparison of visualization test results on SIRST-V2 dataset

Table 2 Ablation experiments to test the performances of various improved methods

编号	模型	mIOU	F₁	P_d	F_a	FLOPs	Parameters
1	YOLOv10n	0.441	0.612	0.56	0.323	6.7G	2.3M
2	YOLOv10n+L-SCDown	0.512	0.677	0.682	0.328	8.5 G	2.9M
3	YOLOv10n+L-SCDown+L-C2f	0.47	0.639	0.679	0.396	5.9 G	2.3M
4	L-YOLOv10n	0.5	0.667	0.738	0.392	5.9G	2.3M

Table 3 Ablation experiments on various L-SCDown variants

Table 4 Complexity analysis and verification of L-SCDown

编号	模型	FLOPs	Parameters
1	SCDown	0.443G	0.036M
2	L-SCDown使用标准卷积	0.877G	0.166M
3	L-SCDown使用空洞卷积	0.877G	0.166M

Table 5 Ablation experiments on various L-C2f variants

Table 6 Complexity analysis and verification of L-C2f

编号	模型	FLOPs	Parameters
1	C2f	1.023G	0.079M
3	L-C2f	0.279G	0.023M

Table 7 Focaler-IOU hyperparameter selection

Fig.10 Performance analysis and experiment of loss function

Table 8 Generalization performance test on NUDT-SIRST dataset

Fig.11 Comparison of visualization test results on NUDT dataset

References 31

[7]	BETTI A, TUCCI M. YOLO-S:a lightweight and accurate YOLO-like network for small target detection in aerial imag-ery[J]. Sensors, 2023, 23(4):1865. doi: 10.3390/s23041865 URL
[8]	CHUNG W Y, LEE I H, PARK C G. Lightweight infrared small target detection network using full-scale skip connec-tion U-Net[J]. IEEE Geoscience and Remote Sensing Letters, 2023, 20:1-5.
[9]	YU C, LIU Y P, ZHAO J M, et al. LR-Net:a lightweight and robust network for infrared small target detection[C] // Proceedings of Conference on Pattern Recognition.Cham, Switzerland: Springer Nature, 2024:19-33.
[10]	YU C, LIU Y P, WU S H, et al. Infrared small target detection based on multiscale local contrast learning networks[J]. Infrared Physics & Technology, 2022, 123:104107.
[11]	KOU R K, WANG C P, PENG Z M, et al. Infrared small target segmentation networks: a survey[J]. Pattern Recognition, 2023, 143:109788. doi: 10.1016/j.patcog.2023.109788 URL
[12]	YUAN S, QIN H L, YAN X, et al. SCTransNet:spatial-channel cross transformer network for infrared small target detection[J]. Transactions on Geoscience and Remote Sensing, 2024, 62:1-15.
[13]	HU C, HUANG Y A, LI K X, et al. DATransNet:dynamic attention transformer network for infrared small target detection[J]. IEEE Geoscience and Remote Sensing Letters, 2025, 22:1-1.
[14]	ZHANG T F, LI L X, CAO S Y, et al. Attention-guided pyramid context networks for detecting infrared small target under complex background[J]. IEEE Transactions on Aerospace and Electronic Systems, 2023, 59(4):4250-4261. doi: 10.1109/TAES.2023.3238703 URL
[15]	ZHAO H Y, ZHANG H P, ZHAO Y Y. YOLOV7-sea:object detec-tion of maritime UAV images based on improved YOLOv7[C]// Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision.Waikoloa,HI,US:IEEE, 2023:233-238.
[16]	WANG Z, HUA Z X, WEN Y C, et al. E-YOLO:recogni-tion of estrus cow based on improved YOLOv8n model[J]. Expert Systems with Applications, 2024, 238:122212. doi: 10.1016/j.eswa.2023.122212 URL
[17]	WANG C Y, YEH I H, LIAO H Y M. YOLOV9:learning what you want to learn using programmable gradient information:arXiv:2402.13616[R]. Ithaca,NY, US: Cornell University, 2024:2402.13616.
[18]	WANG A, CHEN H, LIU L H, et al. Yolov10:real-time end-to-end object detection[J]. Advances in Neural Information Processing Systems, 2024, 37:107984-108011.
[19]	HOU Q Y, WANG Z P, TAN F J, et al. RISTDnet:robust infrared small target detection network[J]. IEEE Geoscience and Remote Sensing Letters, 2021, 19:1-5.
[20]	HU J, SHEN L, ALBANIE S, et al. Squeeze-and-excitation:arXiv:1709.01507[R]. Ithaca,NY, US: Cornell University, 2017:1709.01507.
[21]	LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Venice,Italy:IEEE, 2017:2980-2988.
[22]	LIU Y X, SHEN S J. Vehicle detection and tracking based on improved YOLOv8[J]. IEEE Access, 2025, 13:24793-24803. doi: 10.1109/ACCESS.2025.3538556 URL
[23]	DAI Y M, WU Y Q, ZHOU F, et al. Asymmetric contextual modulation for infrared small target detection[C]// Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision.Waikoloa,HI, US:IEEE, 2021:950-959.
[24]	LI B Y, XIAO C, WANG L G, et al. Dense nested attention network for infrared small target detection[J]. IEEE Tran-sactions on Image Processing, 2022, 32:1745-1758.
[25]	WU X, HONG D F, CHANUSSOT J. UIU-Net:U-Net in U-Net for infrared small object detection[J]. IEEE Transactions on Image Processing, 2022, 32:364-376. doi: 10.1109/TIP.2022.3228497 URL
[26]	ZHAO J M, YU C, SHI Z L, et al. Gradient-guided learning network for infrared small target detection[J]. IEEE Geoscience and Remote Sensing Letters, 2023, 20:1-5.
[1]	熊远辉, 罗中杰, 于光保, 等. 分子Faraday旋光红外滤波成像技术研究[J]. 兵工学报, 2019, 40(7):1443-1449. doi: 10.3969/j.issn.1000-1093.2019.07.014
	XIONG Y H, LUO Z J, YU G B, et al. Research on infrared molecular faraday imaging filter technology[J]. Acta Armamentarii, 2019, 40(7):1443-1449. (in Chinese) doi: 10.3969/j.issn.1000-1093.2019.07.014
[2]	MA T L, YANG Z, LIU B X, et al. A lightweight infrared small target detection network based on target multiscale context[J]. IEEE Geoscience and Remote Sensing Letters, 2022, 20:1-5.
[3]	曾溢良, 蓝金辉, 邹金霖. 滑动置信度约束的红外弱小目标跟踪算法研究[J]. 兵工学报, 2017, 38(9):1771-1778. doi: 10.3969/j.issn.1000-1093.2017.09.014
	ZENG Y L, LAN J H, ZOU J L. Research on infrared dim-small target tracking algorithm with template sliding con-fidence constraint[J]. Acta Armamentarii, 2017, 38(9):1771-1778. (in Chinese)
[4]	王强, 吴乐天, 李红, 等. 基于双支网络协作的红外弱小目标检测[J]. 兵工学报, 2023, 44(10):3165-3176. doi: 10.12382/bgxb.2022.0605
	WANG Q, WU L T, LI H, et al. An infrared small target de-tection method via dual network collaboration[J]. Acta Armamentarii, 2023, 44(10):3165-3176. (in Chinese)
[5]	ZHANG R, LIU M, LI Z. Research on fast detection method of infrared small targets under resource-constrained condi-tions[J]. Journal of Infrared and Millimeter Waves, 2024, 43(4):1-10.
[6]	YU C, LIU Y P, WU S H, et al. Pay attention to local contrast learning networks for infrared small target detection[J]. IEEE Geoscience and Remote Sensing Letters, 2022, 19:1-5.
[27]	MA L, ZHAO L Y, WANG Z X, et al. Detection and counting of small target apples under complicated environments by using improved YOLOv7-tiny[J]. Agronomy, 2023, 13(5):1419. doi: 10.3390/agronomy13051419 URL
[28]	WANG C Y, BOCHKOVSKIY A, LIAO H Y M. YOLOv7:trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.Vancouver,BC, Canada:IEEE, 2023:7464-7475.
[29]	VIJAYAKUMAR A, VAIRAVASUNDARAM S. YOLO-based object detection models:a review and its applications[J]. Multimedia Tools and Applications, 2024, 83 (35):83535-83574. doi: 10.1007/s11042-024-18872-y
[30]	WU M, WANG G C, ZHANG W, et al. Bi-YOLOv10:sample imbalance-aware feature fusion-based object detection for power inspection[C]// Proceedings of the 2024 IEEE First International Conference on Data Intelligence and Innovative Application.Nanning,China:IEEE, 2024:1-6.
[31]	PHAT N T, GIANG N L, DUY B D. GAN-UAV-YOLOv10s:Improved YOLOv10s network for detecting small UAV targets in mountainous conditions based on infrared image data[J]. Neural Computing and Applications, 2025, 37:1-13. doi: 10.1007/s00521-024-10453-2