基于在线增强和跨尺度特征重建的雾天目标检测

doi:10.3778/j.issn.1002-8331.2409-0183

摘要/Abstract

摘要： 针对雾天图像整体亮度和对比度较低且受噪声干扰严重，导致目标检测算法检测效果不佳的问题，提出一种雾天目标检测方法（EC-RTDETR）。设计了在线去雾网络DDNet，利用跳跃连接和轮廓特征计算，去除噪声干扰的同时增强图像的纹理特征；引入轻量化FasterNet作为特征提取网络，充分提取目标空间特征，提高了模型的计算效率；使用加性注意力代替原网络的AIFI，实现更丰富的浅层和深层特征交互；提出了上下文引导的跨尺度特征重建融合模块，设计多尺度网络来提取上下文特征，并对提取后的特征进行融合重构，突出了雾天目标特征，提高了算法在雾天场景中的检测效果。在RTTS数据集上的实验表明：较于基准RTDETR，mAP提高了2.35个百分点，mmAP提高了3.66个百分点，召回率提升了4.71个百分点，参数量减少了9.82×106，证明了EC-RTDETR在实现轻量化的同时，有效提升了雾天场景下的目标检测性能。

关键词: 图像去雾, 目标检测, RTDETR, 跨尺度特征重建, 轻量化网络

Abstract: A foggy target detection method EC-RTDETR is proposed to overcome the problem that the overall contrast of foggy images is low and the images are seriously interfered by noise, which leads to the poor detection effect of target detection algorithm. Firstly, an online defogging network DDNet is designed to remove the noise interference while enhancing the texture information of the image by using jump connections and contour feature computation. Secondly, light-weight FasterNet is introduced as a feature extraction network to fully extract the target spatial features and improve the computational efficiency of the model. Then, efficient additive attention is used instead of the AIFI of the original network to achieve richer shallow and deep feature interaction. Finally, a context-guided cross-scale feature reconstruction fusion module is proposed to enable a multi-scale network design to extract contextual information, and the extracted features are fused and reconstructed to highlight the foggy target features and improve the detection effect of the algorithm in foggy scenes. Experiments on RTTS dataset show that compared with the benchmark RTDETR, mAP is improved by 2.35 percentage points, mmAP is improved by 3.66 percentage points, Recall is improved by 4.71 percentage points, and the amount of parameters is reduced by 9.82×106, which proves that EC-RTDETR effectively improves the performance of target detection in foggy sky scenes while realizing light-weighting.

Key words: image defogging, object detection, RTDETR, cross-scale feature reconstruction, light-weighting network

朱开源, 吴佰靖, 高德勇, 刘媛, 王付祥. 基于在线增强和跨尺度特征重建的雾天目标检测[J]. 计算机工程与应用, 2025, 61(11): 132-143.

ZHU Kaiyuan, WU Baijing, GAO Deyong, LIU Yuan, WANG Fuxiang. Target Detection in Foggy Conditions Based on Online Enhancement and Cross-Scale Feature Reconstruction[J]. Computer Engineering and Applications, 2025, 61(11): 132-143.

参考文献

[1] 郭迎, 梁睿琳, 王润民. 基于CNN图像增强的雾天跨域自适应目标检测[J]. 计算机工程与应用, 2023, 59(16): 187-195.
GUO Y, LIANG R L, WANG R M. Cross-domain adaptive object detection based on CNN image enhancement in foggy conditions[J]. Computer Engineering and Applications, 2023, 59(16): 187-195.
[2] 赖镜安, 陈紫强, 孙宗威, 等. 基于YOLOv5的轻量级雾天目标检测方法[J]. 计算机工程与应用, 2024, 60(6): 78-88.
LAI J A, CHEN Z Q, SUN Z W, et al. Lightweight foggy weather object detection method based on YOLOv5[J]. Computer Engineering and Applications, 2024, 60(6): 78-88.
[3] 万雨昊. 复杂道路环境下的车辆牌照检测与识别[J]. 计算机辅助工程, 2024, 33(2): 31-37.
WAN Y H. Vehicle license plate detection and recognition in complex road environments[J]. Computer Aided Engineering, 2024, 33(2): 31-37.
[4] ZHANG L, WANG M, DING Y, et al. MS-FRCNN: a multi-scale faster RCNN model for small target forest fire detection[J]. Forests, 2023, 14(3): 616.
[5] HUO B, LI C, ZHANG J, et al. SAFF-SSD: self-attention combined feature fusion-based SSD for small object detection in remote sensing[J]. Remote Sensing, 2023, 15(12): 3027.
[6] YU Z, HUANG H, CHEN W, et al. YOLO-Facev2: a scale and occlusion aware face detector[J]. Pattern Recognition, 2024, 155: 110714.
[7] PU Y, LIANG W, HAO Y, et al. Rank-DETR for high quality object detection[J]. arXiv:2310.08854, 2023.
[8] ZHAO C, SUN Y, WANG W, et al. MS-DETR: efficient DETR training with mixed supervision[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024: 17027-17036.
[9] KIM J, LEE M, HEO J P. Self-feedback DETR for temporal action detection[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023: 10286-10296.
[10] ZHANG G, LUO Z, HUANG J, et al. Semantic-aligned matching for enhanced detr convergence and multi-scale feature fusion[J]. arXiv:2207.14172, 2022.
[11] 袁姮, 颜廷昊, 张晟翀. 两阶段特征迁移图像去雾算法[J]计算机工程与应用, 2025, 61(4): 241-252.
YUAN H, YAN T H, ZHANG S Y. Two-stage feature transfer image dehazing algorithm[J]. Computer Engineering and Applications, 2025, 61(4): 241-252.
[12] 吴攀超, 郑卓纹, 王婷婷, 等. 基于CF-YOLO的雾霾交通标志识别[J]. 计算机工程与设计, 2024, 45(7): 2203-2211.
WU P C, ZHENG Z W, WANG T T, et al. Foggy traffic sign recognition based on CF-YOLO[J]. Computer Engineering and Design, 2024, 45(7): 2203-2211.
[13] 刘庆敏, 冯贺阳, 王中, 等. 基于全局及局部优势特征融合的遥感图像去雾方法[J]. 海军航空大学学报, 2024, 39(4): 467-474.
LIU Q M, FENG H Y, WANG Z, et al. Remote sensing image dehazing method based on global and local advantageous feature fusion[J]. Journal of Naval Aviation University, 2024, 39(4): 467-474.
[14] JIN X, TANG R, LIU L, et al. Vehicle license plate recognition for fog‐haze environments[J]. IET Image Processing, 2021, 15(6): 1273-1284.
[15] 吴正平, 岑帅红. 基于自适应gamma校正估计的图像去雾算法[J]. 液晶与显示, 2022, 37(1): 106-115.
WU Z P, CEN S H. Image dehazing algorithm based on adaptive gamma correction estimation[J]. Liquid Crystals and Displays, 2022, 37(1): 106-115.
[16] SAHU G, SEAL A, JAWOREK-KORJAKOWSKA J, et al. Single image dehazing via fusion of multilevel attention network for vision-based measurement applications[J]. IEEE Transactions on Instrumentation and Measurement, 2023, 72: 1-15.
[17] YIN X, YU Z, FEI Z, et al. PE-YOLO: pyramid enhancement network for dark object detection[C]//Proceedings of the International Conference on Artificial Neural Networks, 2023: 163-174.
[18] ZHANG L, ZHAO J, LANG Z, et al. Vehicle detection algorithm for foggy based on improved AOD-Net[J]. Transactions of the Institute of Measurement and Control, 2024, 46 (14): 2696-2705.
[19] ZHENG Y, ZHAN J, HE S, et al. Curricular contrastive regularization for physics-aware single image dehazing[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 5785-5794.
[20] WU R Q, DUAN Z P, GUO C L, et al. RIDCP: revitalizing real image dehazing via high-quality codebook priors[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 22282-22291.
[21] ZHENG Q, SAPONARA S, TIAN X, et al. A real-time constellation image classification method of wireless communication signals based on the lightweight network MobileViT[J]. Cognitive Neurodynamics, 2024, 18(2): 659-671.
[22] ZHENG Q, TIAN X, YU Z, et al. MobileRaT: a lightweight radio transformer method for automatic modulation classification in drone communication systems[J]. Drones, 2023, 7(10): 596.
[23] SINDAGI V A, OZA P, YASARLA R, et al. Prior-based domain adaptive object detection for hazy and rainy conditions[C]//Proceedings of the 16th European Conference on Computer Vision, 2020: 763-780.
[24] 关家志. 基于特征融合的复杂天气下图像识别鲁棒模型[J]. 信息与电脑(理论版), 2022, 34(7): 143-145.
GUAN J Z. Robust image recognition model under complex weather based on feature fusion[J]. Information and Computer (Theoretical Edition), 2022, 34(7): 143-145.
[25] LI J, XU R, MA J, et al. Domain adaptive object detection for autonomous driving under foggy weather[C]//Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2023: 612-622.
[26] 崔雨箫, 陈超, 潘宝峰, 等. 基于深度学习的复杂环境下船舶视觉检测方法[J]. 船舶工程, 2024, 46(2): 106-114.
CUI Y X, CHEN C, PAN B F, et al. Deep learning-based ship detection approach under complex environments[J]. Ship Engineering, 2024, 46(2): 106-114.
[27] HNEWA M, RADHA H. Integrated multiscale domain adaptive YOLO[J]. IEEE Transactions on Image Processing, 2023, 32: 1857-1867.
[28] 王新蕾, 王硕, 翟嘉政, 等. 多任务联合学习下的复杂天气航拍图像目标检测算法[J]. 计算机工程与应用, 2025, 61(2): 97-111.
WANG X L, WANG S, ZHAI J Z, et al. Multi-task joint learning algorithm for aerial image object detection in complex weather[J]. Computer Engineering and Applications, 2025, 61(2): 97-111.
[29] LIU W, REN G, YU R, et al. Image-adaptive YOLO for object detection in adverse weather conditions[C]//Proceedings of the AAAI Conference on Artificial Intelligence. 2022: 1792-1800.
[30] RANJAN R, AVASTHI V. Edge detection using guided sobel image filtering[J]. Wireless Personal Communications, 2023, 132(1): 651-677.
[31] GUO A, SUN K, ZHANG Z. A lightweight YOLOv8 integrating FasterNet for real-time underwater object detection[J]. Journal of Real-Time Image Processing, 2024, 21(2): 49.
[32] SHAKER A, MAAZ M, RASHEED H, et al. SwiftFormer: efficient additive attention for transformer-based real-time mobile vision applications[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023: 17425-17436.
[33] YUE G, LI S, CONG R, et al. Attention-guided pyramid context network for polyp segmentation in colonoscopy images[J]. IEEE Transactions on Instrumentation and Measurement, 2023, 72: 1-13.
[34] NI Z, CHEN X, ZHAI Y, et al. Context-guided spatial feature reconstruction for efficient semantic segmentation[J]. arXiv:2405.06228, 2024.
[35] LI B, REN W, FU D, et al. Benchmarking single-image dehazing and beyond[J]. IEEE Transactions on Image Processing, 2018, 28(1): 492-505.
[36] MENG X, LIU Y, FAN L, et al. YOLOv5s-Fog: an improved model based on YOLOv5s for object detection in foggy weather scenarios[J]. Sensors, 2023, 23(11): 5321.