[1]魏雅丽,牛为华.改进 YOLOv5s 的轻量化航拍小目标检测算法[J].计算机技术与发展,2024,34(02):53-59.[doi:10. 3969 / j. issn. 1673-629X. 2024. 02. 008]
　WEI Ya-li,NIU Wei-hua.Aerial Small Target Detection Based on Improved YOLOv5s Lightweight Algorithm[J].,2024,34(02):53-59.[doi:10. 3969 / j. issn. 1673-629X. 2024. 02. 008]
点击复制

改进 YOLOv5s 的轻量化航拍小目标检测算法()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
34
期数:
2024年02期
页码:
53-59
栏目:
媒体计算
出版日期:
2024-02-10

文章信息/Info

Title:
Aerial Small Target Detection Based on Improved YOLOv5s Lightweight Algorithm
文章编号:
1673-629X(2024)02-0053-07
作者:
魏雅丽; 牛为华
华北电力大学 计算机系,河北 保定 071000
Author(s):
WEI Ya-li; NIU Wei-hua
Department of Computer,North China Electric Power University,Baoding 071000,China
关键词:
小目标检测; 无人机图像; YOLOv5s; 跨层级特征融合; 多尺度检测
Keywords:
small target detection; drone images; YOLOv5s; cross-level feature fusion; multiscale detection
分类号:
TP391. 41
DOI:
10. 3969 / j. issn. 1673-629X. 2024. 02. 008
摘要:
针对无人机航拍图像中小目标样本多、拍摄目标背景复杂、可提取特征信息少的问题,提出一种改进 YOLOv5s 的轻量化无人机航拍小目标检测算法。 首先,改进算法网络结构,增加两条特
征信息传播路径,跨层级连接避免特征损失,同时同级前后连接补充特征信息,并在特征融合过程中加入空间注意力机制,提高模型对小目标区域的关注程度,保留充足的目标特征信息;其次,针
对数据集的特点,将骨干网络中低层小目标检测层融入到特征金字塔网络和路径聚合网络结构中,增加一个检测极小目标的头部;最后,在预测过程中引入 SIoU Loss 定位损失函数,进一步加快模型收敛速度,提升模型检测能力及定位精度。 将该算法在 VisDrone2019 数据集上进行实验,结果表明,改进后的模型 mAP50 达到了 38. 5% ,较基线方法 YOLOv5s 提高了 5. 9 百分点,同时与主流的检测方法相比也取得更高的检测精度,对于小目标检测任务具有较好的性能。
Abstract:
Aiming at the problems of large number of small and medium - sized target samples, complex target background and littleextracted feature information in UAV aerial photography images,an improved YOLOv5s lightweight UAV aerial photography small targetdetection algorithm was proposed. Firstly,the network structure of the algorithm is improved,and two feature information propagationpaths are added to avoid feature loss through cross-level connection. At the same time,the feature information is supplemented by cross-level connection and spatial attention mechanism is added in the feature fusion process to improve the model’s attention to small targetregions and retain sufficient target feature information. Secondly,　according to the characteristics of the data set,the low-level small targetdetection layer of the backbone network is integrated into the feature pyramid network and the path aggregation network structure,and asmall target detection head is added. Finally, SIoU Loss is introduced into the prediction process to further accelerate the modelconvergence speed and improve the model detection ability and positioning accuracy. The proposed algorithm was tested on theVisDrone2019 dataset,which showed that the improved model mAP50 reached 38. 5% ,
5. 9 percentage points higher than that of thebaseline method YOLOv5s, and also achieved higher detection accuracy than the that of mainstream detection methods, with better　performance for small target detection tasks.

相似文献/References:

[1]陶晓力,刘宁钟,沈家全.基于深度信息融合的航拍车辆检测[J].计算机技术与发展,2019,29(09):117.[doi:10. 3969 / j. issn. 1673-629X. 2019. 09. 023]
　TAO Xiao-li,LIU Ning-zhong,SHEN Jia-quan.Aerial Vehicle Detection Based on Depth Information Fusion[J].,2019,29(02):117.[doi:10. 3969 / j. issn. 1673-629X. 2019. 09. 023]
[2]产世兵,刘宁钟,沈家全.基于轻量级网络的 PCB 元器件检测[J].计算机技术与发展,2020,30(10):14.[doi:10. 3969 / j. issn. 1673-629X. 2020. 10. 003]
　CHAN Shi-bing,LIU Ning-zhong,SHEN Jia-quan.PCB Component Detection Based on Lightweight Network[J].,2020,30(02):14.[doi:10. 3969 / j. issn. 1673-629X. 2020. 10. 003]
[3]耿 文,孙 涵.一种 Anchor Free 的无人机检测方法[J].计算机技术与发展,2021,31(01):54.[doi:10. 3969 / j. issn. 1673-629X. 2021. 01. 010]
　GENG Wen,SUN Han.An Anchor Free Method for Detecting Drones[J].,2021,31(02):54.[doi:10. 3969 / j. issn. 1673-629X. 2021. 01. 010]
[4]杨 健,孙 涵.基于深度特征提取的无人机检测算法[J].计算机技术与发展,2021,31(11):71.[doi:10. 3969 / j. issn. 1673-629X. 2021. 11. 012]
　YANG Jian,SUN Han.Drone Detection Based on Depth Feature Extraction[J].,2021,31(02):71.[doi:10. 3969 / j. issn. 1673-629X. 2021. 11. 012]
[5]严 波,任仰勋,方登洲,等.基于串联 Faster R-CNN 的绝缘子自爆检测识别[J].计算机技术与发展,2021,31(12):175.[doi:10. 3969 / j. issn. 1673-629X. 2021. 12. 029]
　YAN Bo,REN Yang-xun,FANG Deng-zhou,et al.Detection and Identification of Insulator Self-explosion Based on Cascaded Faster R-CNN Network[J].,2021,31(02):175.[doi:10. 3969 / j. issn. 1673-629X. 2021. 12. 029]
[6]杨朝晨,陈佳悦,邢 可,等.基于改进的 DSSD 的小目标检测算法研究[J].计算机技术与发展,2022,32(06):63.[doi:10. 3969 / j. issn. 1673-629X. 2022. 06. 011]
　YANG Zhao-chen,CHEN Jia-yue,XING Ke,et al.Small Target Detection Algorithm Based on Improved DSSD[J].,2022,32(02):63.[doi:10. 3969 / j. issn. 1673-629X. 2022. 06. 011]
[7]马田源,孙 涵.改进 RetinaNet 特征融合方式的无人机检测方法[J].计算机技术与发展,2022,32(12):103.[doi:10. 3969 / j. issn. 1673-629X. 2022. 12. 016]
　MA Tian-yuan,SUN Han.An Improved Feature Fusion Method for Drone Detection Based on RetinaNet Extraction[J].,2022,32(02):103.[doi:10. 3969 / j. issn. 1673-629X. 2022. 12. 016]
[8]刘梦菲,毛建华,陆小锋.基于注意力和上下文感知的海面渔桩检测[J].计算机技术与发展,2023,33(08):144.[doi:10. 3969 / j. issn. 1673-629X. 2023. 08. 021]
　LIU Meng-fei,MAO Jian-hua,LU Xiao-feng.Detection of Fishing Piles on Sea Surface Based on Attention and Context Awareness[J].,2023,33(02):144.[doi:10. 3969 / j. issn. 1673-629X. 2023. 08. 021]
[9]文 青,伍 欣,敖 斌,等.基于航空图像的目标检测算法 Trans_YOLOv5[J].计算机技术与发展,2024,34(01):77.[doi:10. 3969 / j. issn. 1673-629X. 2024. 01. 012]
　WEN Qing,WU Xin,AO Bin,et al.Target Detection Algorithm Trans_YOLOv5 Based on Aerial Image[J].,2024,34(02):77.[doi:10. 3969 / j. issn. 1673-629X. 2024. 01. 012]
[10]赵 侃,汪慧兰,郭娇娇,等.基于 DTA-FSAF 的无人机小目标检测研究[J].计算机技术与发展,2024,34(04):101.[doi:10. 3969 / j. issn. 1673-629X. 2024. 04. 016]
　ZHAO Kan,WANG Hui-lan,GUO Jiao-jiao,et al.Research on Small Object Detection of UAV Based on DTA-FSAF[J].,2024,34(02):101.[doi:10. 3969 / j. issn. 1673-629X. 2024. 04. 016]

更新日期/Last Update: 2024-02-10