| 4 | 0 | 3 |
| 下载次数 | 被引频次 | 阅读次数 |
针对空中红外小目标探测存在目标距离远、热辐射衰减严重及环境信息复杂等问题，提出稀疏空间特征与线性注意力融合的对空红外小目标检测（aerial infrared small target detection,AISTD）模型。首先，为解决因特征图尺寸降低而导致目标细节信息丢失的问题，设计了稀疏空间特征提取模块和对称式挤压激励特征提取模块来加强模型对红外小尺度目标的特征提取能力；然后，针对热辐射信号易受环境干扰的问题，将二维图像转换为一维序列后输入到多头线性注意力模块，并对其进行旋转位置编码，从多个注意力子空间中获取红外目标细节纹理和轮廓特征信息，以增强模型在复杂场景下对红外小目标的关注度；最后，建立对空红外小目标数据集并对模型进行训练和测试，以验证模型的有效性。结果表明：所提方法在对空红外小目标数据集上的mAP75达到90.3%,mAP50-95达到74.7%，模型推理时间缩短至3.9 ms，相比现有的YOLOv11和YOLOv13等目标检测网络，AISTD能够更快速准确地识别出空中红外小目标。
Abstract：Aerial infrared small-target detection faces critical challenges including long detection distances, severe thermal radiation attenuation, and complex environmental interference. To address these issues, an aerial infrared small-target detection(AISTD)model that effectively integrates sparse spatial features with linear attention was proposed. First, to alleviate the loss of fine-grained feature information caused by the reduced feature map, a sparse spatial feature extraction module and a symmetrical squeezing excitation feature extraction module were introduced to enhance the model's capability of infrared smallscale target feature extraction. Subsequently, to mitigate environmental interference with thermal radiation signals, the two-dimensional image was converted into a one-dimensional sequence and input into a multihead linear attention module, with rotary positional encoding applied. The proposed model captured detailed texture and contour information of infrared targets across multiple attention subspaces, thereby improving the sensitivity to small targets in complex aerial scenes. Furthermore, an aerial infrared small-target dataset was constructed to train and test the proposed approach and validate its effectiveness. Extensive experimental results demonstrate that the proposed AISTD achieves an mAP75 of 90.3% and an mAP50-95 of 74.7% on a self-built dataset, while reducing the inference time to 3.9 ms. Compared with existing object detection networks, such as YOLOv11 and YOLOv13, the AISTD model exhibits superior performance in terms of both detection accuracy and computational efficiency.
[1]Zhang Yuting,Li Zhengzhou,Siddique A,et al.Infrared small target detection based on interpretation weighted sparse method[J]. IEEE Transactions on Geoscience and Remote Sensing,2025,63:5001415.
[2]Wang Xiaozhen,Han Chengshan, Li Jiaqi, et al.Multiscale feature extraction U-Net for infrared dimand small-target detection[J].Remote Sensing,2024,16(4):643.
[3]姚林杰，王耀斌，蒋伟.巴特沃斯高通滤波器的优化设计与仿真实现[J].机电工程技术，2023,52(2):292-296.Yao Linjie,Wang Yaobin,Jiang Wei.Optimization design and simulation of Butterworth high pass filter[J]. Mechanical&Electrical Engineering Technology,2023,52(2):292-296.
[4]黄西莹，张旭辉，黄玉，等.基于国产异构FPGA的空中红外目标检测跟踪系统[J].红外与激光工程，2024,53(8):114-124.Huang Xiying,Zhang Xuhui,Huang Yu,et al.Domestic heterogeneous FPGA-based aerial infrared target detection and tracking system[J].Infrared and Laser Engineering,2024,53(8):114-124.
[5]贺遵亮，黄营磊.基于TopHat背景建模的运动小目标检测算法分析[J].电子技术，2022,51(8):64-66.He Zunliang, Huang Yinglei. Analysis of moving small target detection algorithm using TopHat background modeling[J]. Electronic Technology, 2022,51(8):64-66.
[6]Gao Chenqiang,Meng Deyu,Yang Yi,et al.Infrared patch-image model for small target detection in a single image[J].IEEE Transactions on Image Processing,2013,22(12):4996-5009.
[7]Dai Yimian,Wu Yiquan.Reweighted infrared patchtensor model with both nonlocal and local priors for single-frame small target detection[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2017, 10(8):3752-3767.
[8]Wang Guanghui,Tao Bingjie,Kong Xuan,et al.Infrared small target detection using nonoverlapping patch spatial-temporal tensor factorization with capped nuclear norm regularization[J].IEEE Transactions on Geoscience and Remote Sensing,2021,60:5001417.
[9]梁锐，魏阳杰.自适应B样条小波函数模糊距离测量方法[J].中国图象图形学报，2021,26(3):503-515.Liang Rui,Wei Yangjie.Adaptive distance measurement method with blur of B-spline wavelet function[J].Journal of Image and Graphics,2021,26(3):503-515.
[10]曾瀚林，孟祥勇，钱惟贤.高斯差分滤波图像融合方法[J].红外与激光工程，2020, 49(S1):168-176.Zeng Hanlin,Meng Xiangyong,Qian Weixian.Image fusion algorithm based on DOG filter[J]. Infrared and Laser Engineering,2020,49(S1):168-176.
[11]Deng He,Sun Xianping,Liu Maili,et al.Entropy-based window selection for detecting dim and small infrared targets[J].Pattern Recognition,2017,61:66-77.
[12]Ren Shaoqing,He Kaiming,Girshick R,et al.Faster R-CNN:towards real-time object detection with region proposal networks[EB/OL].(2016-01-06)[2024-11-29].https：//arxiv.org/abs/1506.01497.
[13]Juhartini,Arwidiyarti D,Desmiwati.Single shot multibox detector(SSD)in object detection:a review[J].IJACI:International Journal of Advanced Computing and Informatics,2025,1(2):118-127.
[14]Hou Qingyu,Wang Zhipeng,Tan Fanjiao, et al.RISTDnet:robust infrared small target detection network[J].IEEE Geoscience and Remote Sensing Letters,2021,19:7000805.
[15]Zhu Ge,Chen Yuhan,Wang Xianquan,et al.MMFF-NET:multi-layer and multi-scale feature fusion network for low-light infrared image enhancement[J]. Signal, Image and Video Processing, 2024,18(2):1089-1097.
[16]Du Xinyue, Cheng Ke, Zhang Jin, et al. Infrared small target detection algorithm based on improved dense nested U-Net network[J]. Sensors,2025,25(3):814.
[17]Mulyana D I,Rowis M A I.Optimization of text mining detection of Tajweed reading laws using the Yolov8 method on the Qur’an[J].QALAMUNA:Jurnal Pendidikan, Sosial, Dan Agama, 2022, 14(2):1089-1110.
[18]Udeh C P,Chen Luefeng,Du Sheng,et al.Improved ShuffleNet V2 network with attention for speech emotion recognition[J]. Information Sciences,2025,689:121488.
[19]Philip Chen C L,Li Hong,Wei Yantao,et al.A local contrast method for small infrared target detection[J]. IEEE Transactions on Geoscience and Remote Sensing,2014,52(1):574-581.
[20]Tang Jingyi,Xu Bu,Li Jue,et al.Ghost-YOLOGBH:a lightweight framework for robust small traffic sign detection via GhostNet and bidirectional multiscale feature fusion[J].Eng,2025,6(8):196.
[21]Bu Yifan, Liu Hao, Li Hongda, et al. Grading algorithm for Orah sorting line based on improved ShuffleNet V2[J].Applied Sciences,2025,15(8):4483.
[22]Lyu Weiyi,Xu Kai,Wu Chonglong,et al.Automatic identification method for subaqueous gravity flow facies in core images using convolutional neural networks[J]. Earth Science Informatics, 2024,18(1):119.
[23]胡伟超，郭宇阳，张奇，等.基于改进YOLOX的轻量化交通监控目标检测算法[J].计算机工程与应用，2024,60(7):167-174.Hu Weichao,Guo Yuyang,Zhang Qi,et al.Lightweight traffic monitoring object detection algorithm based on improved YOLOX[J]. Computer Engineering and Applications,2024,60(7):167-174.
[24]冯伟，薛如翔，傅志凌，等.基于注意力机制和改进YOLOv3的红外弱小目标检测[J].飞控与探测，2023,6(4):84-94.Feng Wei,Xue Ruxiang,Fu Zhiling,et al.Infrared small target detection based on improved YOLOv3 and attention mechanism[J]. Flight Control&Detection,2023,6(4):84-94.
[25]邵延华，张兴平，张晓强，等.联合结构重参数和YOLOv5的航拍红外目标检测[J].电子科技大学学报，2024,53(3):382-389.Shao Yanhua,Zhang Xingping,Zhang Xiaoqiang,et al.Infrared aerial object detection based on structural reparameterization and YOLOv5[J].Journal of University of Electronic Science and Technology of China,2024,53(3):382-389.
[26]许晓阳，高重阳.改进YOLOv7-tiny的轻量级红外车辆目标检测算法[J].计算机工程与应用，2024,60(1):74-83.Xu Xiaoyang,Gao Chongyang.Improved YOLOv7-tiny lightweight infrared vehicle target detection algorithm[J].Computer Engineering and Applications,2024,60(1):74-83.
[27]赵晶钰，李敏，陈谢发，等.融合注意力机制与全局信息的SAR舰船目标检测算法[J].火箭军工程大学学报，2024,38(4):40-46.Zhao Jingyu,Li Min,Chen Xiefa,et al.SAR ship target detection algorithm combining attention mechanism and global information[J].Journal of Rocket Force University of Engineering,2024,38(4):40-46.
基本信息:
DOI：10.20189/j.cnki.CN/61-1527/E.202602005
中图分类号:E91;TP391.41;TN219
引用信息:
[1]陈光晨,张枫,张印辉,等.稀疏空间特征与线性注意力融合的对空红外小目标检测[J].火箭军工程大学学报,2026,40(02):55-65.DOI:10.20189/j.cnki.CN/61-1527/E.202602005.
基金信息:
国家自然科学基金（62561032,62061022,62171206）; 装备智能运用教育部重点实验室开放基金（AAIE-2023-0203）
2026-04-15
2026-04-15