基于轻量级 YOLOv8-ASP 的航空行李检测

Aviation baggage detection based on lightweight YOLOv8-ASP

摘要

航空行李具有种类数量繁多、纹理图案复杂和形状各异等特点，这些特点使得深度学习模型需要更高的计算复杂度才能对行李目标进行准确识别。为了增强工程实用性和降低模型复杂度，该文提出了一种基于轻量级 YOLOv8-ASP 的航空行李检测方法。首先，为优化模型的特征提取和融合能力，同时降低模型复杂性和计算量，设计了自适应 Single PAN 特征融合网络模块；其次，为增强对航空行李重要特征的感知能力，在 YOLOv8n 主干网络的底部加入 ECA 注意力机制；最后，为解决航空行李目标尺度变化引起的不平衡问题，在检测头中引入 MPDIoU 边框损失函数。通过构建航空行李数据集进行实验验证，结果表明，相对于 YOLOv8n 原始模型，所提 YOLOv8-ASP 方法在 mAP@50 上提高了 0.4%，FPS 提高了 16%，参数量从 3.012×10 ⁶ 减少到 2.759×10 ⁶，GFLOPs 从 8.2 减少到 5.6。实验结果表明，所提方法在保证检测精度的同时，显著降低了模型的复杂性和计算量。

Abstract

Aviation luggage detection presents several challenges due to the wide variety of types, complex textures, and varying shapes of luggage, these challenges require deep learning models to have higher computational complexity to accurately identify the luggage targets. To improve the practical applicability and reduce model complexity, this paper proposes an aviation luggage detection method based on the lightweight YOLOv8-ASP. First, to optimize feature extraction and fusion while reducing model complexity and computation, an adaptive single PAN (path aggregation network) feature fusion network module is designed. Secondly, to enhance the perception of important features for aviation luggage, an efficient channel attention (ECA) mechanism is incorporated to the bottom of the YOLOv8n backbone network. Finally, to address the imbalance caused by scale variations of luggage targets, the MPDIoU bounding box loss function is introduced into the detection head. Experimental validation on an aviation baggage dataset demonstrates that, compared to the original YOLOv8n model, the proposed YOLOv8-ASP method achieves a 0.4% improvement in mAP@50, a 16% improvement in FPS, while reducing the parameter count from 3.012×10 ⁶ to 2.759×10 ⁶ and GFLOPs from 8.2 to 5.6. The experimental results show that the proposed method significantly reduces model complexity and computational cost while maintaining detection accuracy.

关键词

航空行李检测 / 轻量化 / YOLOv8 / 自适应特征融合 / 注意力机制

Key words

aviation luggage detection / lightweight / YOLOv8 / adaptive feature fusion / attention mechanism

参考文献

[1] CHEN R B , ZHANG H , LI C , et al. LSDNN: Local—salient deep neural network for baggage re—identification with material discerning[C]// Proceedings of the Chinese Automation Congress. [S.l.]: IEEE, 2020: 6344-6349.

[2] GAO Q J , LIANG P W . Airline baggage appearance transportability detection based on A novel dataset and sequential hierarchical sampling CNN model[J]. IEEE Access, 2021, 9: 41833-41843.

[3] ZHAO Q , MA H M , LU R Q , et al. MVAD—net: Learning view—aware and domain—invariant representation for baggage re—identification[M]// Pattern Recognition and Computer Vision. Cham: Springer International Publishing, 2021: 142-153.

[4] WU H B , LUO Z M , CAO D L , et al. Attention and multi—granied feature learning for baggage re—identification[M]// Computer Supported Cooperative Work and Social Computing. Singapore: Springer, 2022: 460-472.

[5] YANG H , CHU X X , ZHANG L , et al. QuadNet: Quadruplet loss for multi—view learning in baggage re—identification[J]. Pattern Recognition, 2022, 126: 108546.

[6] 邢志伟, 朱书杰, 李彪 . 基于改进图卷积神经网络的航空行李特征感知[J]. 浙江大学学报（工学版）, 2024, 58(5): 941-950.

[7] XING Z W , ZHU S J , LI B . Airline baggage feature perception based on improved graph convolutional neural network[J]. Journal of Zhejiang University (Engineering Science), 2024, 58(5): 941-950.

[8] REDMON J , DIVVALA S , GIRSHICK R , et al. You only look once: Unified, real—time object detection[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2016: 779-788.

[9] REDMON J , FARHADI A . YOLO9000: Better, faster, stronger[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2017: 6517-6525.

[10] REDMON J , FARHADI A . YOLOv3: An incremental improvement[EB/OL]. [ 2025—02—12]. https://arxiv.org/abs/1804.02767.

[11] WANG J X , LIU M , DU Y R , et al. PG—YOLO: An efficient detection algorithm for pomegranate before fruit thinning[J]. Engineering Applications of Artificial Intelligence, 2024, 134: 108700.

[12] JIAO R Z , XU Z M , ZHANG Z H . Improving YOLOv8 via triple attention for pedestrian detection [C]// Proceedings of the 5th International Conference on Computer Vision, Image and Deep Learning. New York: IEEE, 2024: 1517-1521.

[13] CHENG G , CHAO P Z , YANG J , et al. SGST—YOLOv8: An improved lightweight YOLOv8 for real—time target detection for campus surveillance[J]. Applied Sciences, 2024, 14(12): 5341.

[14] LIU S , QI L , QIN H F , et al. Path aggregation network for instance segmentation[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2018: 8759-8768.

[15] WANG Q L , WU B G , ZHU P F , et al. ECA—net: Efficient channel attention for deep convolutional neural networks[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. [S.l.]: IEEE, 2020: 11534-11542.

[16] MA S L , XU Y . MPDIoU: A loss for efficient and accurate bounding box regression[EB/OL]. [ 2025—02—12]. https://arxiv.org/abs/2307.07662.

[17] CHEN J R , KAO S H , HE H , et al. Run, don’t walk: Chasing higher FLOPS for faster neural networks[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2023: 12021-12031.

[18] TAN M X , PANG R M , LE Q V . EfficientDet: Scalable and efficient object detection[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. [S.l.]: IEEE, 2020: 10781-10790.

[19] HU J , SHEN L , SUN G . Squeeze—and—excitation networks[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2018: 7132-7141.

[20] WANG X F , SONG J . ICIoU: Improved loss based on complete intersection over union for bounding box regression[J]. IEEE Access, 2021, 9: 105686-105695.

[21] TAN M , LE Q V . EfficientNet: Rethinking model scaling for convolutional neural networks[EB/OL]. [ 2025—02—12]. https://arxiv.org/abs/1905.11946.

[22] GIRSHICK R . Fast R—CNN[C]// Proceedings of the IEEE International Conference on Computer Vision. New York: IEEE, 2015: 1440-1448.

[23] XU S L , WANG X X , LV W Y , et al. PP—YOLOE: An evolved version of YOLO[EB/OL]. [ 2025—02—12]. https://arxiv.org/abs/2203.16250.

[24] LYU C , ZHANG W , HUANG H , et al. RTMDet: An empirical study of designing real—time object detectors[EB/OL]. [ 2025—02—12]. https://arxiv.org/abs/2212.07784.

[25] LIU W , ANGUELOV D , ERHAN D , et al. SSD: Single shot multiBox detector[C]// Computer Vision—ECCV 2016. Cham: Springer International Publishing, 2016: 21-37.

[26] 刘成润 . 行李分拣系统中的检测与跟踪研究[D]. 成都: 四川大学, 2023.

[27] LIU C R . Research on detection and tracking in baggage sorting system[D]. Chengdu: Sichuan University, 2023.

[28] GE Z , LIU S T , WANG F , et al. YOLOX: Exceeding YOLO series in 2021[EB/OL]. [ 2025—02—12]. https://arxiv.org/abs/2107.08430.

基金资助

四川省科技成果转移转化示范项目(2024ZHCG0044)