Infrared small target detection (ISTD) has important application value in the fields of military early warning, disaster rescue and precision guidance, etc. However, the existing methods have core problems such as an imbalance between local and global features, and insufficient robustness of the model under small sample conditions. To address these issues, this paper proposed a new model called multiscale dynamic fusion with local contrast learning (MDL), which was based on the hybrid architecture of Transformer and convolutional neural network (CNN), integrating multi-scale dynamic weighting and adaptive local contrast constraints. Firstly, a dynamic weighted fusion multi-scale context augmentation (MCA) was introduced into the architecture of Transformer, which effectively solved the problem of imbalance between local and global features in small target detection, and at the same time significantly reduced the redundancy of the attention mechanism. Secondly, this model included a dual-branch hybrid dilated convolution (DHDC) module located in the deeper layers of the network, which enhanced the detection accuracy through high-resolution feature reconstruction and semantic information from deeper features as a way to strengthen the target edge representation. In addition, a local contrast convolution (LCC) module was designed, which incorporated the traditional local contrast idea to significantly reduce the model’s dependence on the number of training samples and maintains excellent robustness even on small sample datasets. Finally, a series of experiments were conducted on NUAA, IRSTD-1k and SIRSTAUG datasets. The results show that the method proposed in this paper achieved F1-Scores of 88.98%, 83.51%, and 86.38% on the three datasets, representing improvements of 1.06, 2.48, and 0.76 percentage points compared with the baseline model, respectively.
针对上述问题, 本文提出了基于多尺度动态融合与局部对比学习的网络模型(Multiscale Dynamic Fusion with Local Contrast Learning for Infrared Small Target Detection, MDLNet), 通过多尺度动态融合与局部对比学习的协同优化机制, 实现了ISTD在特征增强、 噪声抑制与样本效率上的三重突破。
OUYANGY, SHIB, HUANGX, et al. Research on infrared point target recognition method based on space-based early warning system[C]//International Conference on Algorithm, Imaging Processing, and Machine Vision (AIPMV 2023), 2024: 129692U-5.
[2]
CHENGY, LAIX, XIAY, et al. Infrared dim small target detection networks: A review[J]. Sensors, 2024, 24(12): 3885.
[3]
LIS, LIC, YANGX, et al. Infrared dim target detection method inspired by human vision system[J]. Optik, 2020, 206: 164167.
[4]
LIH, WANGQ, WANGH, et al. Infrared small target detection using tensor based least mean square[J]. Computers & Electrical Engineering, 2021, 91: 106994.
[5]
HANJ, MORADIS, FARAMARZII, et al. A local contrast method for infrared small-target detection utilizing a tri-layer window[J]. IEEE Geoscience and Remote Sensing Letters, 2020, 17(10): 1822-1826.
[6]
RENS, HEK, GIRSHICKR, et al. Faster R-CNN: Towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149.
[7]
REDMONJ, DIVVALAS, GIRSHICKR, et al. You only look once: Unified, real-time object detection[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016: 779-788.
[8]
LIUW, ANGUELOVD, ERHAND, et al. SSD: Single shot multibox detector[C]//Computer Vision—ECCV 2016. Cham: Springer, 2016: 21-37.
[9]
DAIY, WUY, ZHOUF, et al. Asymmetric contextual modulation for infrared small target detection[C]//2021 IEEE Winter Conference on Applications of Computer Vision (WACV), 2021: 949-958.
[10]
DAIY, WUY, ZHOUF, et al. Attentional local contrast networks for infrared small target detection[J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 59(11): 9813-9824.
[11]
WANGH, ZHOUL, WANGL. Miss detection vs. false alarm: Adversarial learning for small object segmentation in infrared images[C]//2019 IEEE/CVF International Conference on Computer Vision (ICCV), 2020: 8508-8517.
[12]
LIB, XIAOC, WANGL, et al. Dense nested attention network for infrared small target detection[J]. IEEE Transactions on Image Processing, 2023, 32: 1745-1758.
[13]
KAURR, KAURS. Automatic skin lesion segmentation using attention residual U-Net with improved encoder-decoder architecture[J]. Multimedia Tools and Applications, 2025, 84(8): 4315-4341.
[14]
WANGL, SHENY, ZHANGH, et al. Automatic road extraction framework based on codec network[J]. Journal of Measurement Science and Instrumentation, 2024, 15(3): 318-327.
[15]
HUOX, SUNG, TIANS, et al. HiFuse: Hierarchical multi-scale feature fusion network for medical image classification[J]. Biomedical Signal Processing and Control, 2024, 87: 105534.
[16]
ZHANGY, YUANX, LUOL, et al. Local contrast learning for one-shot learning[J]. Applied Sciences, 2024, 14(12): 5217.
[17]
DAIY, WUY, ZHOUF, et al. Asymmetric contextual modulation for infrared small target detection[C]//2021 IEEE Winter Conference on Applications of Computer Vision (WACV), 2021: 949-958.
[18]
ZHANGT, CAOS, PUT, et al. AGPCNet: Attention-guided pyramid context networks for infrared small target detection[DB/OL]. (2021-11-05) [2025-08-04].
[19]
ZHANGM, ZHANGR, YANGY, et al. ISNet: Shape matters for infrared small target detection[C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022: 867-876.
[20]
FANL, WANGY, HUG, et al. Diffusion-based continuous feature representation for infrared small-dim target detection[J]. IEEE Transactions on Geoscience and Remote Sensing, 2024, 62: 1-17.
[21]
ZHANGM, YANGH, GUOJ, et al. IRPruneDet: Efficient infrared small target detection via wavelet structure-regularized soft channel pruning[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2024, 38(7): 7224-7232.
[22]
PANP, WANGH, WANGC, et al. ABC: Attention with bilinear correlation for infrared small target detection[C]//2023 IEEE International Conference on Multimedia and Expo (ICME), 2023: 2381-2386.
[23]
GAOC, MENGD, YANGY, 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.
[24]
DAIY, WUY. Reweighted infrared patch-tensor 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.
[25]
ZHANGL, PENGZ. Infrared small target detection based on partial sum of the tensor nuclear norm[J]. Remote Sensing, 2019, 11(4): 382.