基于Efficientnet的无人机车辆目标检测算法

doi:10.3778/j.issn.1002-8331.2207-0426

摘要/Abstract

摘要： 针对无人机航拍图像中存在背景复杂，造成车辆漏检严重，检测精度低，以及现有深度网络存在参数量过多、检测速度慢的问题，提出一种基于Efficientnet的无人机车辆目标检测算法。引入轻量化网络Efficientnet作为YOLOv3模型特征提取网络，降低模型参数量，从而提高算法检测速度；采用[K]-means聚类算法对无人机车辆数据集真实框进行聚类，得到更为精确的边界框尺寸，提高检测的精度；使用CIoU边界框损失函数改进模型回归损失，提高模型收敛能力。实验结果表明，改进后的算法在自制无人机车辆数据集中mAP达到92.60%，FPS达到31.15，相对于原始算法分别提高了2.12个百分点和9.87，更加适用于无人机场景下的车辆检测任务。

关键词: 无人机, 车辆检测, YOLOv3, Effcientnet, CIoU

Abstract: In response to the problems of complex background in the aerial images of the UAV which brings about seriously missed detection as well as poor detection precision, and excessive parameters as well as slow detection speed in the present deep network, an object detection algorithm of UAV vehicle based on Eficientnet is proposed. Firstly, Eficientnet, the lightweight network, is introduced as the feature extraction network of the YOLOv3 model, reducing the size of the model parameters and improving the detection speed of the algorithm. Secondly, the UAV vehicle dataset is clustered by the [K]-means algorithm to obtain the better bounding box size, thus improving the detection precision. Finally, with the CIoU bounding box loss function, the regression loss of the model is improved, thus enhancing the convergence ability of the model. Experimental results show that mAP of 92.60% and FPS of 31.15 can be achieved in the self-made UAV vehicle dataset based on the improved algorithm, which is 2.12 percentage points and 9.87 higher than the original algorithm, respectively. Therefore, it is more suitable for vehicle detection tasks in UAV scenarios.

Key words: UAV, vehicle detection, YOLOv3, Effcientnet, CIoU

江德港, 江智, 黄子杰, 郭彩玲, 李柏林. 基于Efficientnet的无人机车辆目标检测算法[J]. 计算机工程与应用, 2023, 59(12): 228-234.

JIANG Degang, JIANG Zhi, HUANG Zijie, GUO Cailing, LI Bailin. UAV Vehicle Object Detection Algorithm Based on Efficientnet[J]. Computer Engineering and Applications, 2023, 59(12): 228-234.

参考文献

[1] DALAL N，TRIGGS B.Histograms of oriented gradients for human detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，2005：886-893.
[2] LUO J，OUBONG G.A comparison of SIFT，PCA-SIFT and SURF[J].International Journal of Image Processing，2009，3（4）：143-152.
[3] FELZENSZWALB P F，GIRSHICK R B，MCALLESTER D，et al.Object detection with discriminatively trained part-based models[J].IEEE Transactions on Pattern Analysis and Machine Intelligence，2009，32（9）：1627-1645.
[4] GIRSHICK R，DONAHUE J，DARRELL T，et al.Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，2014：580-587.
[5] GIRSHICK R.Fast R-CNN[C]//Proceedings of the IEEE International Conference on Computer Vision，2015：1440-1448.
[6] REN S，HE K，GIRSHICK R，et al.Faster R-CNN：towards real-time object detection with region proposal networks[C]//Advances in Neural Information Processing Systems，2015：91-99.
[7] REDMON J，DIVVALA S，GIRSHICK R，et al.You only look once：unified，real time object detection[C]//IEEE Conference on Computer Vision and Pattern Recognition，2016.
[8] REDMON J，FARHADI A.YOLO 9000：better，faster，stronger[C]//IEEE Conference on Computer Vision and Pattern Recognition，2017.
[9] REDMON J，FARHADI A.Yolov3：an incremental improvement[C]//IEEE Conference on Computer Vision and Pattern Recognition，2018.
[10] BOCHKOVSKIY A，WANG C Y，LIAO H Y M.YOLOv4：optimal speed and accuracy of objectdetection[J].arXiv：2004.10934，2020.
[11] LIU W，ANGUELOV D，ERHAN D，et al.SSD：single shot multi box detector[C]//European Conference on Computer Vision.Cham：Springer，2016：21-37.
[12] LIN T Y，GOYAL P，GIRSHICK R，et al.Focal loss for dense object detection[J].IEEE Transactions on Pattern Analysis & Machine Intelligence，2017，99：2999-3007.
[13] 袁小平，马绪起，刘赛.改进YOLOv3的行人车辆目标检测算法[J].科学技术与工程，2021，21（8）：3192-3198.
YUAN X P，MA X Q，LIU S.An improved algorithm of pedestrian and vehicle detection based on YOLOv3[J].Science Technology and Engineering，2021，21（8）：3192-3198.
[14] 宋欢欢，惠飞，景首才.改进的RetinaNet模型的车辆目标检测[J].计算机工程与应用，2019，55（13）：225-230.
SONG H H，HUI F，JING S C.Improved RetinaNet model for vehicle target detection[J].Computer Engineering and Applications，2019，55（13）：225-230.
[15] 钱惠敏，陈纬，马宜龙，等.基于ResNet34＿D改进YOLOv3模型的行人检测算法[J].控制与决策，2022，37（7）：1713-1720.
QIAN H M，CHEN W，MA Y L，et al.Pedestrian detection based on the developed YOLOv3 with ResNet34_D[J].Control and Decision，2022，5（7）：1713-1720.
[16] TAN M，LE Q V.EfficientNet：rethinking model scaling for convolutional neuralnetworks[C]//Proceedings of the 36th International Conference on Machine Learning，2019：6105-6114.
[17] 叶佳林，苏子毅，马浩炎，等.改进YOLOv3的非机动车检测与识别方法[J].计算机工程与应用，2021，57（1）：194-199.
YE J L，SU Z Y，MA H Y，et al.Improved YOLOv3 non-motor vehicles detection and recognition method[J].Computer Engineering and Applications，2021，57（1）：194-199.
[18] 宋艳艳，谭励，马子豪，等.改进YOLOV3算法的视频目标检测[J].计算机科学与探索，2021，15（1）：163-172.
SONG Y Y，TAN L，MA Z H，et al.Video target detection based on improved YOLOV3 algorithm[J].Journal of Frontiers of Computer Science and Technology，2021，15（1）：163-172.
[19] 徐涛，陈奕仁，吕宗磊.基于改进YOLOv3的机坪工作人员反光背心检测研究[J].计算机科学，2022，49（4）：239-246.
XU T，CHEN Y R，LV Z L.Study on reflective vest detection for apron workers based on improved YOLOv3 algorithm[J].Computer Science，2022，49（4）：239-246.
[20] REZATOFIGHI H，TSOI N，GWAK J Y，et al.Generalized intersection over union：a metric and a loss forbounding box regression[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，2019：658-666.
[21] ZHENG Z H，WANG P，LIU W，et al.Distance-IOU loss：faster and better learning for bounding box regression[C]//Proceedings of AAAI Conference on Artificial Intelligence，2020.