改进SSD算法的道路小目标检测研究

doi:10.19678/j.issn.1000-3428.0061499

摘要/Abstract

摘要： 在道路场景中，因小目标分辨率低且特征不明显，传统的目标检测算法难以确认其所属类别和位置信息，导致检测精度低、检测速度慢、漏检率高。提出一种改进SSD的道路小目标检测算法RFG_SSD。在SSD网络结构的主干部分和检测部分之间，通过引入改进的特征金字塔网络结构，融合浅层和深层感受野的特征信息，以获得小目标语义信息丰富的特征图。将深层特征提取网络ResNet 50作为改进网络的主干特征提取网络，提高整体网络的检测精度。为加快网络运算速度，基于检测层结构，利用全局平均池化层代替全连接层，减少网络参数量。实验结果表明，与SSD、VGG16+SFPN等算法相比，该算法能够有效提高小目标检测性能，且加快检测速度，其在BDD100K数据集上的平均精度和检测速度分别为98.05%和85.56 frame/s，小目标检测个数相较于SSD算法提高3倍多。

关键词: 小目标检测, SSD算法, ResNet50网络, 特征金字塔网络, 全局平均池化

Abstract: In a road scene, the resolution of small targets is low, and the characteristics are not obvious. Traditional target detection algorithms find it difficult to identify category and location information, resulting in low detection accuracy, slow detection speed, and a high missed detection rate.This study proposes a small target detection algorithm (RFG_SSD) based on an improved SSD.By introducing the improved Feature Pyramid Network(FPN) structure and integrating the feature information of shallow and deep receptive fields, which obtain a feature image with rich semantic information of small targets, to improve the performance of small target detection.The deep feature extraction network ResNet50 is used as the backbone of the improved network to improve detection accuracy of the overall network.To further accelerate the network operation speed based on the detection layer structure, the Global Average Pooling(GAP) layer is used to replace the full connection layer to reduce the number of network parameters. The experimental results on the BDD100K dataset show that compared with the SSD and VGG16+SFPN algorithms, this algorithm can effectively improve the performance of small-target detection and increase the detection speed.Its average accuracy and detection speed are 98.05% and 85.56 frame/s, respectively.The number of small target detections is more than three times higher than that of the SSD algorithm.

Key words: small target detection, SSD algorithm, ResNet50 network, Feature Pyramid Network(FPN), Global Average Pooling(GAP)

中图分类号:

TP391

邹慧海, 侯进. 改进SSD算法的道路小目标检测研究[J]. 计算机工程, 2022, 48(5): 281-288.

ZOU Huihai, HOU Jin. Research on Road Small Target Detection with Improved SSD Algorithm[J]. Computer Engineering, 2022, 48(5): 281-288.

https://www.ecice06.com/CN/Y2022/V48/I5/281

图/表 11

20220806175018

20220806175022

20220806175026

20220806175029

20220806175033

20220806175037

20220806175040

20220806175049

20220806175053

20220806175057

20220806175100

参考文献

[1] DAVID F.Object detection with discriminatively trained part-based models[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2014, 47(7):6-7.
[2] 段仲静, 李少波, 胡建军, 等.深度学习目标检测方法及其主流框架综述[J].激光与光电子学进展, 2020, 57(12):1-16. DUAN Z J, LI S B, HU J J, et al.Summary of deep learning target detection methods and mainstream frameworks[J].Laser and Optoelectronics Progress, 2020, 57(12):1-16.(in Chinese)
[3] GIRSHICK R, DONAHUE J, DARRELL T, et al.Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2014:580-587.
[4] GIRSHICK R.Fast R-CNN[C]//Proceedings of IEEE International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2015:1440-1448.
[5] REN S, HE K, GIRSHICK R, et al.Faster R-CNN:towards real-time object detection with region proposal networks[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 39(6):1137-1149.
[6] HE K, GKIOXARI G, DOLLÁR D, et al.Mask R-CNN[C]//Proceedings of IEEE International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2017:1-10.
[7] REDMON J, DIVVALA S, GIRSHICK R, et al.You only look once:unified, real-time object detection[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2016:779-788.
[8] REDMON J, FARHADI A.YOLO9000:better, faster, stronger[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2017:6517-6525.
[9] JOSEPH R, FARHADI A.YOLOv3:an incremental improvement[EB/OL].[2021-03-25].https://arxiv.org/pdf/1804.02767.pdf.
[10] BOCHKOVSKIY A, WANG C Y, LIAO H Y M.YOLOv4:optimal speed and accuracy of object detection[EB/OL].[2021-03-25].https://arxiv.org/abs/2004.10934.
[11] LIU W, ANGUELOV D, ERHAN D, et al.SSD:single shot multibox detector[C]//Proceedings of European Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2016:21-37.
[12] TAN M X, PANG R M, LE Q V.EfficientDet:scalable and efficient object detection[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2020:1-10.
[13] WANG C Y, LIAO H Y M, WU Y H, et al.CSPNet:a new backbone that can enhance learning capability of CNN[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, Workshops.Washington D.C., USA:IEEE Press, 2020:390-399.
[14] LIU S, QI L, QIN H F, et al.Path aggregation network for instance segmentation[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2018:1-10.
[15] HE K M, ZHANG X Y, REN S Q, et al.Deep residual learning for image recognition[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2016:1-10.
[16] 田强, 贾小宁.基于深度残差网络的车标识别[J].吉林大学学报(理学版), 2021, 59(2):319-324. TIAN Q, JIA X N.Vehicle logo recognition basedon deep residual network[J].Journal of Jilin University (Science Edition), 2021, 59(2):319-324.(in Chinese)
[17] LI Y Z, YUAN Y.Convergence analysis of two-layer neural networks with ReLU activation[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems.New York, USA:ACM Press, 2017:597-607.
[18] XU J, LI Z S, DU B, et al.Reluplex made more practical:Leaky ReLU[C]//Proceedings of IEEE Symposium on Computers and Communications.Washington D.C., USA:IEEE Press, 2020:1-7.
[19] 刘建伟, 赵会丹, 罗雄麟, 等.深度学习批归一化及其相关算法研究进展[J].自动化学报, 2020, 46(6):1090-1120. LIU J W, ZHAO H D, LUO X L, et al.Research progress on batch normalization of deep learning and its related algorithms[J].Acta Automatica Sinica, 2020, 46(6):1090-1120.(in Chinese)
[20] HSIAO T Y, CHANG Y C, CHOU H H, et al.Filter-based deep-compression with global average pooling for convolutional networks[J].Journal of Systems Architecture, 2019, 95(12):9-18.
[21] LIU W Y, WEN Y D, YU Z D, et al.Large-margin softmax loss for convolutional neural networks[EB/OL].[2021-03-25].https://arxiv.org/pdf/1612.02295.pdf.
[22] YU F, CHEN H F, WANG X, et al.BDD100K:a diverse driving dataset for heterogeneous multitask learning[EB/OL].[2021-03-25].https://arxiv.org/pdf/1805.04687.pdf.
[23] 史文旭, 谭代伦, 鲍胜利.特征增强SSD算法及其在遥感目标检测中的应用[J].光子学报, 2020, 49(1):154-163. SHI W X, TAN D L, BAO S L.Feature enhancement SSD algorithm and its application in remote sensing images target detection[J].Acta Photonnica Sinica, 2020, 49(1):154-163.(in Chinese)
[24] 雷华迪, 陈东方, 王晓峰.基于级联SSD的目标检测算法[J].计算机工程与设计, 2020, 41(12):3517-3524. LEI H D, CHEN D F, WANG X F, Target detection algorithm based on cascaded SSD[J].Computer Engineering and Design, 2020, 41(12):3517-3524.(in Chinese)

选择文件类型/文献管理软件名称

选择包含的内容