基于改进faster RCNN的木材运输车辆检测

doi:10.11772/j.issn.1001-9081.2019081431

摘要/Abstract

摘要： 针对目前森林资源受到盗砍盗伐威胁，相关木材运输车辆行为隐蔽，进而导致无法准确地在交通视频中被识别的问题，提出了一种基于改进fater区域卷积神经网络（faster RCNN）的木材运输车辆检测方法。首先，采用faster RCNN作为基础检测框架，使用金字塔特征网络(FPN)、多尺度训练、锚点框聚类作为基础改进措施；其次，以广义交并比(GIoU)损失函数替换原算法中的smoothL1损失函数作为边界框定位回归的损失函数；最后，计算出在多种实验条件下的模型平均精度均值(mAP),对各种算法进行了对比。实验结果表明，使用GIoU作为损失函数的faster RCNN相比原算法对木材运输车辆检测的平均精度(AP)上升了7.5%，模型平均精度均值（mAP）上升了4.3%;同时，在大型数据集PASCALVOC上，使用GIoU作为损失函数的faster RCNN的mAP达到73.4%，相比其他算法具有明显优势。

关键词: 广义交并比, 目标检测, 损失函数, 金字塔特征网络, faster区域卷积神经网络, 车型检测

Abstract: Because the problem that the current forest resources are threatened by theft and the related timber transport vehicles’s activities are hidden, these vehicles can not be accurately identified in the traffic video. A timber transport vehicle detection method based on improved faster RCNN (Regional Convolutional Neural Network) was proposed. Firstly, faster RCNN was used as the basic detection framework, and Pyramid Feature Network(PFN), multi-scale training, anchor frame clustering were adopted as the basic improvement measures. Secondly, the smoothL1 loss in the original algorithm was replaced with Generalized Intersection over Union(GIoU) loss function as the loss function of the boundary box regression. Finally, the mean Average Precisions(mAPs) of the model under various experimental conditions were calculated, and the various algorithms are compared. The experimental results show that the Average Precision(AP) of the faster RCNN with GIoU as the loss function is 7.5% higher than that of the original algorithm, and the model mAP is increased by 4.3%. At the same time, a comparative experiment was designed on PASCALVOC dataset. The faster RCNN with GIoU loss function has a mAP value of 73.4%, which has significant advantages over the comparison algorithms.

Key words: Generalized Intersection over Union (GIoU), target detection, loss function, Pyramid Feature Network (PFN), faster RCNN (faster Regional Convolutional Neural Network), vehicle detection

中图分类号:

TP391

徐义鎏贺鹏任东王慧董婷邵攀. 基于改进faster RCNN的木材运输车辆检测[J]. 计算机应用, DOI: 10.11772/j.issn.1001-9081.2019081431.

XU Yiliu，HE Peng，REN Dong，WANG Hui，DONG Ting，SHAO Pan. Timeber transport vehicle detection based on improved faster RCNN[J]. Journal of Computer Applications, DOI: 10.11772/j.issn.1001-9081.2019081431.

参考文献

[1] LIU W, ANGUELOV D, ERHAN D, et al.SSD: Single Shot multibox Detector[C]// Proceedings of the 2016 European Conference on Computer Vision, LNCS 9905. Cham: Springer, 2016: 21-37.., :-.(.., ():-.(in chinese))
[2] Joseph Redmon, Santosh Divvala, Ross Girshick, et al.You Only Look Once: Unified, Real-Time Object Detection[C]// Proceedings of the 2016 Conference on Computer Vision and Pattern Recognition.Washington, DC: IEEE Computer Society, 2016: 1-6.., :-.(.., ():-.(in chinese))
[3]. REN S， HE K， GIRSHICK R， et al． Faster R-CNN: towards realtime object detection with region proposal networks[J] IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149．[J]., :-.(.[J]., ():-.(in chinese))
[4] LI Suhao, LIN Jinzhao, LI Guoquan, et al.Vehicle type detection based on deep learning in traffic sceneProcedia[J].Computer Science, 2018, 131: 564-572.[J]., :-.(.[J]., ():-.(in chinese))
[5] SIMONYAN K, ZISSERMAN A.Very Deep Convolutional Networks for Large-Scale Image Recognition[EB/OL]. [2015-04-10]. https://arxiv.org/pdf/1409.1556.pdf.[J]., :-.(.[J]., ():-.(in chinese))
[6] He Kaiming, Zhang Xiangyu, Ren Shaoqing, et al.Deep Residual Learning for Image Recognition[C]// Proceeding of the 2016 IEEE in conference on computer vision and pattern recognition. Piscataway:IEEE, 2016: 770-778.., :-.(.., ():-.(in chinese))
[7] Zheng Qin, Zhaoning Zhang, Xiaotao Chen, et al.Fd-Mobilenet: Improved Mobilenet with a Fast Downsampling Strategy[C]// Proceedings of the 2018 IEEE International Conference on Image Processing. Piscataway:IEEE, 2018: 1363-1367.., :-.(.., ():-.(in chinese))
[8] Y.Shi, Y. Li, X. Wei and Y. Zhou, " A Faster-RCNN Based Chemical Fiber Paper Tube Defect Detection Method[C]// Proceedings of the 2017 5th International Conference on Enterprise Systems. 2017:173-177.doi: 10.1109/ES.2017.35., :-.(.., ():-.(in chinese))
[9] Lin T Y, Dollar P, Girshick R, He Kaiming, et al.Feature Pyramid Networks for Object Detection[C]// Proceedings of the 2017 IEEE in conference on computer vision and pattern recognition. Piscataway:IEEE, 2017: 936-944.., :-.(.., ():-.(in chinese))
[10] REZATOFIGHI H, TSOI N, GWAK J Y, et al.Generalized intersection over union: A metric and a loss for bounding box regression[C]// Proceedings of the 2019 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2019: 658-666.., :-.(.., ():-.(in chinese))
[11] SUN X, WU P, STEVEN C, et al.iFace detection using deep learning: an improved faster RCNN approach[J]. Neurocomputing, 2018, 99: 42-50[J]., :-.(.[J]., ():-.(in chinese))
[12] YU X, YUAN Y.Hand gesture recognition based on Faster-RCNN deep learning[J]. Journal of Computers, 2019, 14(2): 101-110.[J]., :-.(.[J]., ():-.(in chinese))
[13] ZHAO X, LI W, ZHANG Y, GULLIVER T A, et al.A faster RCNN-based pedestrian detection system [C]// Proceedings of the 2016 IEEE 84th Vehicular Technology Conference.Piscataway, NJ: IEEE, 2016:1-5.., :-.(.., ():-.(in chinese))
[14] GU Y, WANG B, XU B.A FPN-based framework for vehicle detection in aerial images[C]// Proceedings of the 2018 International Conference on Video and Image Processing.New York: ACM, 2018: 60-64., :-.(.., ():-.(in chinese))
[15] 唐泽坤, 朱泽宇, 杨裔, 等.基于距离和密度的d-K-means算法[J].计算机应用研究, 2019, 37(6): 1-7.[J]., :-.(.[J]., ():-.(in chinese))
[16] 戴陈卡, 李毅.基于faster RCNN以及多部件结合的机场场面静态飞机检测[J].计算机应用, 2017, 37(S2):85-88.[J]., :-.(.[J]., ():-.(in chinese))
[17] 周玉金, 王晓东, 张力戈, 等.基于卷积神经网络的左心室超声图像特征点定位[J].计算机应用, 2019, 39(4):1201-1207.[J]., :-.(.[J]., ():-.(in chinese))
[18] 邱津怡, 罗俊, 李秀, 等.基于卷积神经网络的多尺度葡萄图像识别方法[J].计算机应用, 2019, 39(10):2930-2936.[J]., :-.(.[J]., ():-.(in chinese))
[19] ZAMIR A R, SAX A, WILLIAM B, et al.Taskonomy: disentangling task transfer learning[C]// Proceedings of the 2018 IEEE Conference on Computer Vision and Pattern Recognitio. Washington, DC: IEEE Computer Society, 2018: 3712-3722., :-.(.., ():-.(in chinese))
[20] LIN T Y, GOYAL P, GIRSHICK R, et al.Focal loss for dense object detection[C]// Proceedings of the 2017 IEEE international conference on computer vision. Piscataway: IEEE, 2017: 2980-2988., :-.(.., ():-.(in chinese))
[21] REDMON J, FARHADI A.YOLO9000: better, faster, stronger[C]// Proceedings of the 2017 IEEE conference on computer vision and pattern recognition. Piscataway:IEEE, 7263-7271.., :-.(.., ():-.(in chinese))